BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:Gathering of 1000+ practitioners from the data ecosystem X-WR-CALDESC:Gathering of 1000+ practitioners from the data ecosystem NAME:The Fifth Elephant 2019 X-WR-CALNAME:The Fifth Elephant 2019 REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:The Fifth Elephant 2019 TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Introduction to The Fifth Elephant 2019 DTSTART:20190725T034500Z DTEND:20190725T040000Z DTSTAMP:20260421T132223Z UID:session/NMw6ufLMhH8r2ZJVwVA7Re@hasgeek.com SEQUENCE:0 CREATED:20190515T061827Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T090805Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to The Fifth Elephant 2019 in Auditorium 1 in 5 m inutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to sessions in Audi 2 DTSTART:20190725T034500Z DTEND:20190725T035500Z DTSTAMP:20260421T132223Z UID:session/S56yz8eCTkvg4C7DjWJTzq@hasgeek.com SEQUENCE:0 CREATED:20190717T044542Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190723T024450Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to sessions in Audi 2 in Auditorium 2 in 5 minute s TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:ADAM - Bootstrapping a deep NN-based sequence labeling model with minimal labeling DTSTART:20190725T035500Z DTEND:20190725T044000Z DTSTAMP:20260421T132223Z UID:session/Lj3Evp2qDHNPgs6QadTZSz@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk,Intermediate,Lecture,Intermediate,Lecture,Full talk o f 40 mins,Scheduled rehearsal,Has potential. Mentor proposer. CREATED:20190617T050250Z DESCRIPTION:We would be presenting answers to the following... \nWhy just any generic approach would not have worked?\nHow our data source and struc ture left us with no previously adapted choices? \nWhy we the project was necessary to meet the end goals of the company?\nAnd how did we tackle a n umber of problems on the way?\n\n###**Company’s Goals**:\nFollowing are the primary product-goals of the company which are relevant to ADAM projec t.:\n1)Universal Product Catalog\n2)Aggregation and Market Analysis\n3)Sel f evolving Knowledge Graph\n\n###**Raw Dataset**:\nIntroduction\, Structur e and The good\, bad and the ugly of the data-set.\n\n\n###**WHY ADAM (Aut omatic Detection and Annotation Module)... A deep NN model**:\nDefinition\ ,\nUsecases::\n1) Enrichment of Knowledge graph\n2) For Analytics\n\n###** Components of ADAM**:\n1)Smart Automatic Training Data Generation\n2)State of the Art Sequence Tagging Model\n3)Active Learning approach\n\n###**Why the above architecture is chosen**:\n1)Zero ground truth and no training data available whatsoever.\n2)Multi-Independent Source of data generation thus imagine the variance \n3)Short representations and extremely noisy\n4 )Prone to Extreme human error (not bias but error!!!)\n\n##**Finally detai ls of the architecture and WHY they were necessary**:\n###**1)Smart Autom atic Training Data Generation**:\n<>How we leveraged the structure of data set (Stock-item and Stock-group)?\n<>How we used the existing knowledge-ba se AKA (CREGS)?\n<>How we improvised using the information from other sour ces like Amazon and GS1?\n\n###**2)State of the Art Sequence Tagging Model **:\n<>Why we created our own word embeddings and how it helped us?\n<>Why BiLSTM and CRF were used and why are they state of the art?\n<>Why this s pecific architecture was needed and why anything else wouldn’t work\n<>W hat was the accuracy and how well did the model performed?\n\n###**3)Activ e Learning**:\n<>Why Active learning when we can generate labels automatic ally?\n<>How we integrated and designed Manual annotation model ourselves? \n<>How well did we reach maturity with the minimum data-points manually l abelled?\n<>Why extrinsic sampling or intrinsic sampling used?\n \n### **Conclusion that we will showcase as per 5th Elephant**:\n1)How to tackle the noisy data problem in case of textual data?\n2)Why a deep NN model pl ays an important role in generalisation?\n3)Why Active Learning is a reall y important concept for dealing with the problem of no label data?\n\n### Speaker bio\n\nIIT Roorkee Grad. (Batch 2017)\nData Scientist (Exp: 1.8 yr s at Clustr\, Tally Analytics pvt. ltd.)\nI have been a part of the Data S cience team at Clustr.\nI have worked on some innovative projects which in volved skills on Deep-Learning\, Machine Learning\, complex data-structure and dynamic programming algorithms.\nI have been the primary owner of ADA M project and have successfully converted it from a problem statement to w orking solution. I brain-stromed\, coded and tackled all the problems face d while the journey of ADAM.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115152Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/adam-bootstrapping-a-d eep-nn-based-sequence-labeling-model-with-minimal-labeling-Lj3Evp2qDHNPgs6 QadTZSz BEGIN:VALARM ACTION:display DESCRIPTION:ADAM - Bootstrapping a deep NN-based sequence labeling model w ith minimal labeling in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Solving the vehicle routing problem for optimizing shipment delive ry DTSTART:20190725T040000Z DTEND:20190725T044000Z DTSTAMP:20260421T132223Z UID:session/FbzNHwetjzuh7SAKtJjZyi@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Full talk of 40 mins CREATED:20190529T061004Z DESCRIPTION:1. Description of the context at the Flipkart delivery hub.\n2 . Solution constraints - customer time window\, maximum number of shipmen ts per vehicle.\n3. Defining a non-standard cost function - total travel t ime\, route outliers\, compactness of routes.\n4. Formulating the problem as a variant of VRPTW (vehicle routing problem with time windows).\n5. Com putational complexity: NP-hard (generalization of Traveling Salesman Probl em)\n6. Overview of some exact algorithms.\n7. Heuristics - (a) constructi on of routes and (b) route improvement\n8. Description of our construction step and iterative computational procedure to improve solutions. \n9. Dis cussion of results\n\n### Speaker bio\n\nVenkateshan Kannan is a data scie ntist with the Logistics and Insight team at Flipkart. With a PhD. in stat istical physics and postdoc in systems biology\, Venkateshan has worked on problems spanning multiple domains in academia and industry. He enjoys ap proaching problems from first principles.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115146Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/formulation-and-soluti on-of-vehicle-routing-problem-for-optimizing-shipment-delivery-routes-FbzN Hwetjzuh7SAKtJjZyi BEGIN:VALARM ACTION:display DESCRIPTION:Solving the vehicle routing problem for optimizing shipment de livery in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Ghostbusters: optimizing debt collections with survival models DTSTART:20190725T044000Z DTEND:20190725T052500Z DTSTAMP:20260421T132223Z UID:session/KXGG7kvX7CiPeid6gDfj92@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Full talk of 40 mins,Confirmed,Has potential. Ment or proposer. CREATED:20190531T074023Z DESCRIPTION:### TL\;DR\nThis talk is about using survival models to optimi ze the process of making collection calls (“dear sir\, please pay your b ill\, it’s overdue”). \n\n### Context:\n* An overview of how the call ing process is structured. This will give an understanding of what we’re trying to optimize.\n* Discuss why moving people from one level of the pr ocess to another automatically and optimally is important for recovering m oney.\n* Get an understanding of why data-backed decisions are important f or overall efficiency. Is it worth it to make 7 calls per user or should y ou escalate after 4 calls?\n* Understand how using panel data for user beh avior is significantly different from more standard classifiers which use cross-sectional data.\n\n### A brief introduction to survival models:\n* W hat survival models are\, and where they are traditionally used. Get an in troduction to basic terminology like survival function\, hazard rate\, cen soring\, etc.\n* Take a look at non-traditional applications of survival m odels in fields like sales lead prioritization\, marketing automation\, et c.\n\n### How we use survival models:\n* How math concepts are directly re levant to the business - a hazard function is directly useful as a lead sc ore\, while a survival function tells us who the ghosts are. Math => busin ess decisions.\n* Constructing hazard curves via parametric (Weibull) and non-parametric (Kaplan-Meier) and connecting them to our real data. \n* Co x proportional model\n* Data limitations force us to use censored models.\ n* Take a look at productionizing these models\; how to use this informati on to make better decisions. One model can solve many problems (escalation \, lead scoring\, write-off\, etc.)\n\n### Speaker bio\n\nFasih is a data scientist at Simpl\, India’s top pay later platform. When he’s not bus y playing video games\, he’s busy writing about all-things-Bayes and fun ctional programming. Prefers adrak-wali-chai over coffee\, suggests orderi ng from Tata Cha over Chai Point\, and paying using Simpl.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115135Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/ghostbusters-optimizin g-debt-collections-with-survival-models-KXGG7kvX7CiPeid6gDfj92 BEGIN:VALARM ACTION:display DESCRIPTION:Ghostbusters: optimizing debt collections with survival models in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Tutorial on core concepts in Social Network Analysis (SNA) DTSTART:20190725T044000Z DTEND:20190725T061000Z DTSTAMP:20260421T132223Z UID:session/DTMgzRyaNYMQM5XNz7faAN@hasgeek.com SEQUENCE:2 CATEGORIES:Tutorial,Tutorial CREATED:20190712T082517Z DESCRIPTION:The outline of this tutorial is as follows: \n\n1. Subject int roduction and motivation\n2. Key concepts and terminology\n3. Network Meas ures\n4. Tools\, software used\n5. Analytical techniques\n\n### Speaker bi o\n\nI am a researcher and data scientist. My research interests are healt hcare\, e-commerce and social media. There are exciting possibilities\, in teresting insights waiting to be uncovered by network analysis. The talk i s wide in its coverage and applications to ensure everyone\, from geeks to politicians or doctors or journalists are kept onboard and learn somethin g new as much as add to others.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/demystifying-social-ne twork-analysis-sna-a-tutorial-DTMgzRyaNYMQM5XNz7faAN BEGIN:VALARM ACTION:display DESCRIPTION:Tutorial on core concepts in Social Network Analysis (SNA) in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Invite-only round table on "Ground truths and platform engineering " by Flipkart's Fulfilment and Services Group (FSG) DTSTART:20190725T045000Z DTEND:20190725T060500Z DTSTAMP:20260421T132223Z UID:session/Df9xcwXMPixwVrJLSrtd1u@hasgeek.com SEQUENCE:1 CREATED:20190717T050200Z DESCRIPTION:1. Introduction to FSG (5 minutes)\n2. Topic 1 - Introduction (15 minutes)\n3. Topic 1 - Open discussion (50 minutes)\n4. Topic 2 - Intr oduction (15 minutes)\n5. Topic 2 - Open discussion (50 minutes)\n6. Closi ng remarks (5 minutes)\n\nProposed Topics:\n\n1. Blue collar workforce man agement: we will discuss the challenges associated with functional modelli ng of this area and related technical challenges.\n\n2. Supply planning fo r last mile: we will discuss the challenges associated with functional mod elling of this area and related technical challenges.\n\n3. Fulfillment pl anning and execution amidst fast changing supply and demand patterns: we w ill discuss the challenges associated with understanding customer preferen ces\, and reacting to supply/demand shocks and exceptions at large volumes .\n\n4. Warehouse layout management: we will discuss the challenges associ ated with modeling the layout of a warehouse and keeping it in sync with c hanges on the ground.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230108T103046Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Invite-only round table on "Ground truths and platform enginee ring" by Flipkart's Fulfilment and Services Group (FSG) in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190725T052500Z DTEND:20190725T055500Z DTSTAMP:20260421T132223Z UID:session/P2V5oM3aEajiH4B6aLjfv6@hasgeek.com SEQUENCE:0 CREATED:20190515T062835Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T050235Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:10 steps to build your own data pipeline from day one of your sta rtup DTSTART:20190725T055500Z DTEND:20190725T063000Z DTSTAMP:20260421T132223Z UID:session/WkaSjk8cqgP1zAwR2skKk7@hasgeek.com SEQUENCE:2 CATEGORIES:Beginner,Full talk of 40 mins,Scheduled rehearsal CREATED:20190529T060906Z DESCRIPTION:1. Be clear of Requirements and Constraints\n - Having a scala ble system for data ingestion\n - Data design (Specific or Generic)\n - Qu erying interface - why stick to SQL?\n2. Take time to Design Data\n - Walk ing through example of generic table design\n3. Sort out Data production p art first\n - Identify all possible data producers (and understand require ments). In our case - \n - Android/iOS app\n - Cannot keep sending each e vent over network\n - Cannot lose data even if app crashes or is killed\n - Keep out of context from the application itself\n - Microservice(s)\n - Cannot keep sending each event over network\n - Keep data collection a gnostic of microservice itself\n4. Design v1.0 of Data pipeline\n - How an d why we chose "anti-pattern"\n5. Choose/Design Data warehouse\n - Data de sign in Redshift\n - Compression ON for certain columns\n - Tuning for sca le\n - Taking care of Querying patterns of Product Managers and Data scien tists\n6. Open up: Enable many Data Interfaces\n - On demand Data loading and querying: OnDemand Table(s)\n - Flexibility for complicated analysis: Adhoc redshift cluster(s)\n7. Understand\, Tune & Repeat\n8. Optimize for Usage\n - Added more columns at generic level e.g.\n - More examples\n9. O ptimize for Cost & Ops\n - Retention policies of data\n - Not all events are of same importance\n - But all events should be accessible if require d\n10. Upgrade to v2.0 of Data pipeline\n\n### Speaker bio\n\nI am Kumar P uspesh\, CTO and Co-Founder of Moonfrog\, India’s top mobile gaming comp any. We had to design a large scale data infrastructre from day 1 of our c ompany to cater to our product needs. Having a cost sensitive as well as s calable approach helped us achieve large scale as a gaming company in Indi a in short amount of time. At the same time taught us a lot of ingenious w ays of building large scale infra customized for business and its users (r ather than a generic paid solution and then changing your usage/requiremen ts based on that).\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/10-steps-to-build-your -own-data-pipeline-for-day-1-of-your-startup-WkaSjk8cqgP1zAwR2skKk7 BEGIN:VALARM ACTION:display DESCRIPTION:10 steps to build your own data pipeline from day one of your startup in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190725T060500Z DTEND:20190725T063500Z DTSTAMP:20260421T132223Z UID:session/KpfuG9cep94yxwaBkXMpTH@hasgeek.com SEQUENCE:0 CREATED:20190717T050307Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T085253Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190725T061000Z DTEND:20190725T064000Z DTSTAMP:20260421T132223Z UID:session/CNsjdiMazDsQbZDa2eUsoH@hasgeek.com SEQUENCE:0 CREATED:20190717T050248Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T053041Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Sponsored talk: Age of AI Ops DTSTART:20190725T063000Z DTEND:20190725T071000Z DTSTAMP:20260421T132223Z UID:session/Kc13fb57CVeDrXBH4u8zJg@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk of 40 mins,Full talk of 40 mins CREATED:20190717T032807Z DESCRIPTION:In this session\, we go over\n* Key attributes of AIOps\n* Tec hnology layers \n* Problems that AIOps promises to address and case studie s\n* Reference architecture\n* Road ahead\n\n### Speaker bio\n\nNitin Gupt a works at Appdynamics\, leading the AppDynamics Data Platform from Bangal ore. Earlier\, he spent 10 years at Microsoft\, had a short stint with his own startup and then led Data and User platforms at Inmobi technologies. He has been working with large scale distributed data processing systems o ver the last 8 years.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115127Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/age-of-ai-ops-Kc13fb57 CVeDrXBH4u8zJg BEGIN:VALARM ACTION:display DESCRIPTION:Sponsored talk: Age of AI Ops in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Invite-only round table on "Ground truths and platform engineering " by Flipkart's Fulfillment and Supply Group (FSG) DTSTART:20190725T063500Z DTEND:20190725T075000Z DTSTAMP:20260421T132223Z UID:session/5hpQS1DWZqVEsefYb2Jv8c@hasgeek.com SEQUENCE:1 CREATED:20190717T050335Z DESCRIPTION:1. Introduction to FSG (5 minutes)\n2. Topic 1 - Introduction (15 minutes)\n3. Topic 1 - Open discussion (50 minutes)\n4. Topic 2 - Intr oduction (15 minutes)\n5. Topic 2 - Open discussion (50 minutes)\n6. Closi ng remarks (5 minutes)\n\nProposed Topics:\n\n1. Blue collar workforce man agement: we will discuss the challenges associated with functional modelli ng of this area and related technical challenges.\n\n2. Supply planning fo r last mile: we will discuss the challenges associated with functional mod elling of this area and related technical challenges.\n\n3. Fulfillment pl anning and execution amidst fast changing supply and demand patterns: we w ill discuss the challenges associated with understanding customer preferen ces\, and reacting to supply/demand shocks and exceptions at large volumes .\n\n4. Warehouse layout management: we will discuss the challenges associ ated with modeling the layout of a warehouse and keeping it in sync with c hanges on the ground.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230108T103046Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Invite-only round table on "Ground truths and platform enginee ring" by Flipkart's Fulfillment and Supply Group (FSG) in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Taking deep learning to production with RedisAI DTSTART:20190725T064000Z DTEND:20190725T072500Z DTSTAMP:20260421T132223Z UID:session/NsrfM3F3uCyGtTbL5cXmB7@hasgeek.com SEQUENCE:2 CREATED:20190722T090117Z DESCRIPTION:Year 2018 was the year of model servers. There were numeroius initiatives for building a reliable\, interoperable deep learning deployme nt toolkits but so far we don't have an easy tool that can reliably handle the deep learning models from all the frameworks. With the advent of Redi s modules and the availability of C APIs for the major deep learning frame works\, it is now possible to turn Redis into a reliable runtime for deep learning workloads\, providing a simple solution for a model serving micro service. In this talk we will introduce RedisAI\, a joint effort by [tenso r]werk and RedisLabs that introduces tensors and graphs as new Redis data types and allows to execute graphs over tensors using multiple backends (P yTorch\, TensorFlow\, and ONNXRuntime)\, both on the CPU and GPU. The modu le also supports scripting with TorchScript\, which provides a Python-like tensor language that can be used to facilitate pre- and post-processing o perations\, like input shaping or output ensembling. In addition\, thanks to its support for the ONNX standard\, including ONNX-ML\, RedisAI is not strictly limited to deep learning\, but it offers support for general mach ine learning algorithms. In this talk\, we will demonstrate a full journey from training a model to deploying to production in a highly available en vironment. Last\, we will lay down the roadmap for the future\, like autom ated batching\, sharding\, integration with Redis data types (e.g. streams ) and advanced monitoring. The talk will include sample code\, best practi ces and a live demo.\n\n### Speaker bio\n\nI am working as a part of the d evelopment team of tensorwerk\, an infrastructure development company focu sing on deep learning deployment problems. I and my team focus on building open source tools for setting up a seamless deep learning workflow. I hav e been programming since 2012 and started using python since 2014 and move d to deep learning in 2015. I am an open source enthusiast and I spend mos t of my research time on improving interpretability of AI models using [Tu ringNetwork](https://turingnetwork.ai). I am part of the core development team of Hangar and RedisAI and a constant contributor to PyTorch source. I also have authored a [deep learning book](https://github.com/hhsecond/Han dsOnDeepLearningWithPytorch). I go by hhsecond on internet\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Taking deep learning to production with RedisAI in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Anatomy of a production ML feature engineering platform DTSTART:20190725T071000Z DTEND:20190725T075000Z DTSTAMP:20260421T132223Z UID:session/88vKTH52YdCQYRhrawExJj@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Full talk of 40 mins,Scheduled rehearsal,Has poten tial. Mentor proposer. CREATED:20190531T073936Z DESCRIPTION:Rough Outline:\n\n+ Objectives of a feature engineering platfo rm (5 mins)\n - Reduce time to market \n - Enhance robustness of mod els \n - Enable explainability\n\n+ Points of friction & required capab ilities (20 mins)\n - What is in my data? (catalog) \n - Is my input data complete and correct? (health)\n - How do I link existing side in formation (augment/enrich)\n - How to capture tacit knowledge/signal (l abeling) \n - How do I reliably prepare my training datasets (pipelines )\n - How do I check audit & validate what has been computed (audit)\n - How do I discover what is being computed and used? (marketplace) \n - How do I export and track exported discovered features for model dev ( search)\n - How do I link the features to performance? (monitor)\n - How do I reuse the features in the streaming path? (library) \n \n+ Ec onomics of Feature Engineering (5 mins)\n - Feature computation expensi ve\, and each has a price\n - Amortization happens over time & across m odels \n - Process discipline required \n - Questions to ask:\n 1. How many models will I have over time? \n 2. How defensible shoul d they be?\n 3. How available should they be?\n 4. How many feat ures will they need?\n\n+ Approaches to building one (5 mins)\n - FEAST (Go-JEK\; Thought through but tied to GCP) \n - Combine standalone com ponents (OSS exists but incur integration costs) \n - Thirdparty (Move fast but incur platform costs)\n\n### Speaker bio\n\nDr. Venkata Pingali i s Co-Founder and CEO of Scribble Data\, an ML Engineering company based in Bangalore and Denver. Scribble’s flagship enterprise product\, Enrich\, accelerates ML productionization in enterprises. Before starting Scribble Data\, Dr. Pingali was VP of Analytics at a political data consulting fir m. He has a BTech from IIT Mumbai and a PhD from USC in Computer Science.\ n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115117Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/anatomy-of-a-productio n-ml-feature-engineering-platform-88vKTH52YdCQYRhrawExJj BEGIN:VALARM ACTION:display DESCRIPTION:Anatomy of a production ML feature engineering platform in Aud itorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190725T072500Z DTEND:20190725T082500Z DTSTAMP:20260421T132223Z UID:session/VFHSmdoqAsNCJ68Av7ceST@hasgeek.com SEQUENCE:0 CREATED:20190717T050545Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T053053Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190725T075000Z DTEND:20190725T083000Z DTSTAMP:20260421T132223Z UID:session/Lecp8VKTHK7NgsKwa2VqYp@hasgeek.com SEQUENCE:0 CREATED:20190717T050628Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T085314Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190725T075000Z DTEND:20190725T085000Z DTSTAMP:20260421T132223Z UID:session/8ye76Q2ut9ymZVxQCpGKWu@hasgeek.com SEQUENCE:0 CREATED:20190515T062054Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190719T071258Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Tutorial: Taking deep learning to production with RedisAI DTSTART:20190725T082500Z DTEND:20190725T095500Z DTSTAMP:20260421T132223Z UID:session/P2zU4UtBNfyHjvtZxF7DF7@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk,Intermediate,Demo,Tutorial CREATED:20190711T103706Z DESCRIPTION:Year 2018 was the year of model servers. There were numeroius initiatives for building a reliable\, interoperable deep learning deployme nt toolkits but so far we don't have an easy tool that can reliably handle the deep learning models from all the frameworks. With the advent of Redi s modules and the availability of C APIs for the major deep learning frame works\, it is now possible to turn Redis into a reliable runtime for deep learning workloads\, providing a simple solution for a model serving micro service. In this talk we will introduce RedisAI\, a joint effort by [tenso r]werk and RedisLabs that introduces tensors and graphs as new Redis data types and allows to execute graphs over tensors using multiple backends (P yTorch\, TensorFlow\, and ONNXRuntime)\, both on the CPU and GPU. The modu le also supports scripting with TorchScript\, which provides a Python-like tensor language that can be used to facilitate pre- and post-processing o perations\, like input shaping or output ensembling. In addition\, thanks to its support for the ONNX standard\, including ONNX-ML\, RedisAI is not strictly limited to deep learning\, but it offers support for general mach ine learning algorithms. In this talk\, we will demonstrate a full journey from training a model to deploying to production in a highly available en vironment. Last\, we will lay down the roadmap for the future\, like autom ated batching\, sharding\, integration with Redis data types (e.g. streams ) and advanced monitoring. The talk will include sample code\, best practi ces and a live demo.\n\n### Speaker bio\n\nI am working as a part of the d evelopment team of tensorwerk\, an infrastructure development company focu sing on deep learning deployment problems. I and my team focus on building open source tools for setting up a seamless deep learning workflow. I hav e been programming since 2012 and started using python since 2014 and move d to deep learning in 2015. I am an open source enthusiast and I spend mos t of my research time on improving interpretability of AI models using [Tu ringNetwork](https://turingnetwork.ai). I am part of the core development team of Hangar and RedisAI and a constant contributor to PyTorch source. I also have authored a [deep learning book](https://github.com/hhsecond/Han dsOnDeepLearningWithPytorch). I go by hhsecond on internet\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115108Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/taking-deep-learning-t o-production-with-redisai-P2zU4UtBNfyHjvtZxF7DF7 BEGIN:VALARM ACTION:display DESCRIPTION:Tutorial: Taking deep learning to production with RedisAI in A uditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: Tackling the complex inter-depende nt challenges in transport planning and assignment DTSTART:20190725T083000Z DTEND:20190725T093000Z DTSTAMP:20260421T132223Z UID:session/QwRWejSMgs58cJXYp1vBm7@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190717T071516Z DESCRIPTION:**Drivers/Discussants:**\n\nAs the focus here is on the techni cal discussion\, the drivers would ideally be the ones who’ve worked on designing the automated system for shipment delivery/route optimization an d/or implementing the solution in the real-world.\n\n**Key takeaways from this session**\n\nLearning about the preferred tools and techniques that a re being employed to solve routing problems. Is there such a thing as stat e-of-the-art in this case? Understanding the similarities and differences in problem formulation\, issues and performance across the different busi nesses/firms. Where is the greatest scope for improvement? Who can bring f resh insights in this area?\n\n### Speaker bio\n\n- Venkateshan K (Flipkar t)\n- Vaibhav Khandelwal (Shadowfax)\n- Jayaram Kasi (Pikkol)\n- Rahul Jai n (Locus)\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/bof-tackling-the-compl ex-inter-dependent-challenges-in-transport-planning-and-assignment-QwRWejS Mgs58cJXYp1vBm7 BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: Tackling the complex inter-dep endent challenges in transport planning and assignment in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Joint Q&A on data pipelines\, operations and scaling challenges DTSTART:20190725T085000Z DTEND:20190725T090500Z DTSTAMP:20260421T132223Z UID:session/DhHcv5DFPNSKpczwixV1vH@hasgeek.com SEQUENCE:0 CREATED:20190717T054035Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T054059Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Joint Q&A on data pipelines\, operations and scaling challenge s in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:The Anaconda journey: challenges faced in building an OSS business with data DTSTART:20190725T090500Z DTEND:20190725T095000Z DTSTAMP:20260421T132223Z UID:session/KR5EQmnzuVK4XB9Kocg9kk@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk of 40 mins CREATED:20190627T082716Z DESCRIPTION:* The Early Years\n * Founding visions\n * State of Pyth on and Scipy in early 2010s\n * Python & "Big Data"\n * Creation of Con da\, Anaconda\, PyData\n * Community\n * Technical initiatives\n * Creating an OSS Business\n * Future\n * Challenges as we grow & scale \n * Technical and Community hurdles\n\n### Speaker bio\n\nPeter Wang i s a co-founder of Anaconda\, Inc.\, where he is CTO and leads the Open Sou rce and Community Innovation team. He has been developing commercial scien tific computing and visualization software for 20 years. He has extensive experience in software design and development across a broad range of area s\, including 3D graphics\, geophysics\, large data simulation and visuali zation\, financial risk modeling\, and medical imaging.\n\nAs a creator of the PyData community and conferences\, he devotes time and energy to grow ing the Python data science community and advocating and teaching Python a t conferences around the world. Peter holds a BA in Physics from Cornell U niversity.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/the-anaconda-journey-K R5EQmnzuVK4XB9Kocg9kk BEGIN:VALARM ACTION:display DESCRIPTION:The Anaconda journey: challenges faced in building an OSS busi ness with data in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of a Feather (BOF) session: Creating a data-driven culture i n the startup ecosystem DTSTART:20190725T093000Z DTEND:20190725T103000Z DTSTAMP:20260421T132223Z UID:session/8KLVhDsBQeo427AKDxPmZz@hasgeek.com SEQUENCE:2 CATEGORIES:Birds of a Feather session of 1 hour CREATED:20190717T072210Z DESCRIPTION:Learn how data driven culture can be inculcated when starting up.\n\n### Speaker bio\n\n- Akash Khandelwal (Flipkart)\n- Goda Ramkumar ( xto10x)\n- Kumar Puspesh (Moonfrog Labs)\n- Venkata Pingali (Scribble Data )\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/birds-of-a-feather-dat a-driven-culture-in-the-startup-ecosystem-8KLVhDsBQeo427AKDxPmZz BEGIN:VALARM ACTION:display DESCRIPTION:Birds of a Feather (BOF) session: Creating a data-driven cultu re in the startup ecosystem in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:The final stage of grief (about bad data) is acceptance DTSTART:20190725T095000Z DTEND:20190725T103000Z DTSTAMP:20260421T132223Z UID:session/CqMnamYKtUJvQbbdoNYfaC@hasgeek.com SEQUENCE:3 CATEGORIES:Advanced,Full talk of 40 mins,Confirmed,Strong accept CREATED:20190530T061707Z DESCRIPTION:Over the course of my career I’ve gone through the many stag es of grief\; I’ve become angry at the poor quality of my data\, I’ve attempted to bargain with engineering/PMs/etc for better data\, and I beca me depressed over the issue. Now I’ve reached the final stage\; I accept that my data is bad. Given that my data is bad\, I then attempt to model it’s badness\, and use that model to correct for the biases introduced.\ n\nIn this talk I’ll discuss how I approach bad data\; I accept that I c annot fix it and instead try to model where it came from. This usually inv olves getting a more detailed grasp of the data generating process and wri ting down a formal model.\n\nIn many cases this enables me to use the data model to correct and enhance my predictive model\, as well as provide use ful measurements and insights for improving and repairing the data collect ion process.\n\n### Speaker bio\n\nChris is currently the head of data sci ence at Simpl\, India's top Pay Later platform. In past lives he's been a physicist\, a high frequency stock trader\, an automated marketer\, a body guard and a nootropic drug courier. He's a strong believer in correct stat istics\, clean code\, and putting skin in the game to demonstrate your bel iefs.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115056Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/it-works-on-the-traini ng-data-is-the-new-it-works-on-my-machine-why-data-science-is-terrible-eng ineering-CqMnamYKtUJvQbbdoNYfaC BEGIN:VALARM ACTION:display DESCRIPTION:The final stage of grief (about bad data) is acceptance in Aud itorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190725T095500Z DTEND:20190725T102500Z DTSTAMP:20260421T132223Z UID:session/JcQABwNnhPiiKN6bMP99JT@hasgeek.com SEQUENCE:0 CREATED:20190717T053450Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T053500Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: Building fulfillment platforms -- India's eCommerce landscape DTSTART:20190725T102500Z DTEND:20190725T112500Z DTSTAMP:20260421T132223Z UID:session/PAuyUVnFrrSymiWfrQXmDb@hasgeek.com SEQUENCE:1 CREATED:20190722T084835Z DESCRIPTION:We will discuss the challenges associated with building platfo rms which need to be in sync with a fast changing physical world while rea cting to shocks and exceptions on the ground.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230108T103046Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: Building fulfillment platforms -- India's eCommerce landscape in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Joint Q&A on bad data and failures in data management DTSTART:20190725T103000Z DTEND:20190725T104500Z DTSTAMP:20260421T132223Z UID:session/2eU6io5aB4zh5igFLY5JGs@hasgeek.com SEQUENCE:0 CREATED:20190717T054238Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T054243Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Joint Q&A on bad data and failures in data management in Audit orium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190725T103000Z DTEND:20190725T110000Z DTSTAMP:20260421T132223Z UID:session/PkyhK2cKXrAWbo9sMC7TuF@hasgeek.com SEQUENCE:0 CREATED:20190717T071540Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T071622Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190725T104500Z DTEND:20190725T111500Z DTSTAMP:20260421T132223Z UID:session/4qyetZskzNXhNLTPc9C2vk@hasgeek.com SEQUENCE:0 CREATED:20190515T062300Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190719T071310Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On ML platforms DTSTART:20190725T110000Z DTEND:20190725T120000Z DTSTAMP:20260421T132223Z UID:session/F2kb4EXRQCCVz7oXhJHWxJ@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190717T071308Z DESCRIPTION:The purpose of this BoF is to have a conversation around platf orms that \norganizations are building develop and deploy ML models. We wi ll discuss a \nnumber of practical challenges in developing and deploying ML Platforms\n\nWe will touch upon :\n\n(a) Whether organizations need one and when? \n(b) What should it achieve? What is it value proposition? \n( c) What is it relationship to cloud offerings such Azure ML? \n(d) How sho uld one go about developing one? \n(e) How should one think about technolo gy/other choices? \n(f) What the challenges in developing and operating on e?\n\nSpecifically we will discuss\n\n(a) Data flows - stability\, scaling \, changing requirements \n(b) Team structure/skill requirements and avail ability\n(c) Development Support - Notebooks\, production vs test\, realti me vs batch \n(d) Life cycle management - Planning\, deployment\, evolutio n \n(e) Operations - Monitoring\, debugging\, evolution to latest tooling\ n(f) Pressures - Balance of need to deliver vs need to architecture \n(g) Processes - For development efficiency\, correctness \n(h) Data Governance - access and data copy management\, privacy\n(i) Scaling - how to grow wi th data sets\, number of models\, computational requirements\, diversity?\ n\n### Speaker bio\n\n- Venkata Pingali\n- Jaidev Deshpande\n- Pramod Bili giri\n- Krupal Modi\n- Ravi SK\n- Govind Pandey\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/bof-on-ml-platforms-F2 kb4EXRQCCVz7oXhJHWxJ BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On ML platforms in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Contracts\, schema evolution and data pipelines DTSTART:20190725T111500Z DTEND:20190725T114500Z DTSTAMP:20260421T132223Z UID:session/41B6R9MQBc5a2aCwzxFoiu@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Short talk of 20 mins,Confirmed,Has potential. Men tor proposer. CREATED:20190531T072329Z DESCRIPTION:The flow would look like this\n\n- The Need for a Message Bus in building a data processing pipeline\n- For the events generated in the Message Bus\, the need for a contract for data control (with examples of s howing how we messed up and learnt from it). \n- - explain in more detail of what a contract is\n- - how it can be implemented\n- - - starts with hi erarchical modeling of data. relations between objects\n- - - what are too ls other there to store this complex relationship between entites\n\n- Dis cuss the gains from implementing contract control for any data that flows in the data pipeline \n- - from a business perspective of improving busine ss logic\, joining with other data sets\n- - from a technological ease - \ n* Schema extendibility of fields in data\, \n* predictability of develop ment\, \n* back dated processing - backward and forward compatibility\n* A ble to break down pipeline by responsibility - teams can work on different component of the pipeline - Implementing the above for multi step data pr ocessing (enrichment)\n\nAdditional Advantages\n* Cost wise\n* Data cleani ng\n* Data consistency\n* Linear pipeline\n\n### Speaker bio\n\nI work as a Technology Architect at zapr. Working closely with data engineering team s and more specifically drive initiatives to help improve the quality of t he data. In my spare time i like to read about how lot of different organi zations are solving new type of problems\, listen to lot of podcasts and w atch football\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115045Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/building-robust-reliab le-data-pipelines-41B6R9MQBc5a2aCwzxFoiu BEGIN:VALARM ACTION:display DESCRIPTION:Contracts\, schema evolution and data pipelines in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Flash talks: by participants DTSTART:20190725T112500Z DTEND:20190725T120000Z DTSTAMP:20260421T132223Z UID:session/SdP22BzFjUFpchPWiwcQxp@hasgeek.com SEQUENCE:0 CREATED:20190717T033759Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T140931Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Flash talks: by participants in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Analysing high throughput data in real-time DTSTART:20190725T114500Z DTEND:20190725T121500Z DTSTAMP:20260421T132223Z UID:session/ACVTgcHN7K4RB6E7z6ECGr@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk of 40 mins,Short talk of 20 mins CREATED:20190703T090103Z DESCRIPTION:- Introduction\n- About Hotstar\n- Stream Processing @Hotstar\ n - What is Stream Processing and Why was it required\n - Problems that lead to usage\n - Video Player Metricing\n - Social Signals\n - U ser Targeting\n- Case Study - Video Player Metrics\n - What are the P1 me trics\n - How did we solve and compute them real time\n- Case Study - Soc ial Signals\n - What are the Social Signals\n - How did we solve engagem ent in real time\n- Key Take Away Discussion\n- Why and when should we use Stream processing\n- Q&A\n\n### Speaker bio\n\nCurrently\, Data Engineeri ng at Hotstar. Previously at WebEngage and co-founder at CareODrive. Inter ested in spreading/sharing knowledge and in solving problems at a scale th at matters. Previously held talks at Golang Meet-Ups\, Bangalore\, India a nd the 21CF Global Data Summit. Big fan of Radiohead\, hit me up for a jam session any time.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115037Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/analysing-high-through put-data-in-real-time-ACVTgcHN7K4RB6E7z6ECGr BEGIN:VALARM ACTION:display DESCRIPTION:Analysing high throughput data in real-time in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to sessions in Audi 2 DTSTART:20190726T035000Z DTEND:20190726T040000Z DTSTAMP:20260421T132223Z UID:session/V7BYpZqUyeTeMoLV2taQWt@hasgeek.com SEQUENCE:0 CREATED:20190717T052228Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T092240Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to sessions in Audi 2 in Auditorium 2 in 5 minute s TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introduction to Day 2\; recap of Day 1 DTSTART:20190726T035000Z DTEND:20190726T040000Z DTSTAMP:20260421T132223Z UID:session/GXdYRnR8owX385HyhQxp4M@hasgeek.com SEQUENCE:0 CREATED:20190710T074038Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T091607Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introduction to Day 2\; recap of Day 1 in Auditorium 1 in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Leveraging the power of analytics for MarTech DTSTART:20190726T040000Z DTEND:20190726T044500Z DTSTAMP:20260421T132223Z UID:session/7GYSLDrX7K3VerYyZUorrC@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Very company specific: maybe,Scheduled rehearsal t o make a decision,Full talk of 40 mins,Submit revised slides within 7 days ,Reject: needs lot of work CREATED:20190704T123300Z DESCRIPTION:- Brief about CleverTap \n- Current State of the Industry \n- Challenges faced by Marketers on Segmentation\n- CleverTap's solution for Intelligent Segmentation with Machine Learning\n- Case studies showing the real impact of CleverTap's solution\n- Challenges faced by Marketers on C ampaign Content\n- CleverTap's solution to Automating Campaign Content wit h Recommender Engine\n- Case studies showing the real impact of CleverTap' s Recommender Engine\n\n### Speaker bio\n\nAn Investment Banker by acciden t and a Data Scientist by choice\, Jacob has successfully transitioned fro m the world of finance to the world of nerds in analytics. Jacob brings ov er 15 years of combined experience in analytics\, consulting\, portfolio m anagement to solve complex business problems faced by marketers with the h elp of cutting-edge industry-first solutions powered by Data Science.\n\nJ acob is a proud recipient of 40 under 40 Data Scientists (2019) awarded by Analytics India Magazine. He has written multiple articles on data scienc e which have been picked up by renowned sites like kdnuggets\, datascience central.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115028Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/leveraging-power-of-an alytics-for-martech-7GYSLDrX7K3VerYyZUorrC BEGIN:VALARM ACTION:display DESCRIPTION:Leveraging the power of analytics for MarTech in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Similarity search for product matching at Semantics3 DTSTART:20190726T040000Z DTEND:20190726T044000Z DTSTAMP:20260421T132223Z UID:session/Qx9qKgrCf9d3N5qNJjjzMt@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Lecture,Full talk of 40 mins,Submit revised slides within 7 days,Has potential. Mentor proposer. CREATED:20190711T103521Z DESCRIPTION:### Introduction [~ 5 mins]\n\nThis section will present an ov erview of the problem\, the use cases that motivate it and establish the t one for the rest of the presentation.\n\nTopics Covered:\n\n- Product Matc hing: What is it and Why is it important?\n- Similarity Search for Product Matching: What is it and how does it speed up matching?\n- Example Case f or Similarity Search: Sample product document and sample query document to explain the following sections.\n\n---\n\n### Traditional Text Search App roaches [~ 5 mins]\n\nThis section will cover our intial attempt at the si milarity search problem using traditional text based methods largely lever aging elasticsearch.\n\nTopics Covered:\n\n- Overview of how we set up the problem\n- Bottlenecks we hit and available tuning options\n- Examples of real queries \n\n\n### Lessons from Traditional Text Search Approaches [~ 5 mins]\n\nThis section will cover some of the key insights we gleaned fr om traditional text approaches and how we needed to reframe the problem.\n \nTopics Covered:\n\n- The nature of our data/problem and why elasticsearc h wasn't a good fit.\n- Need for indexing multi-modal data\n- Examples of failed cases\n- Search is only as good as the document's representation.\n \n---\n\n### Representation Learning [~ 10 mins]\n\nThis section would cov er how we reframed this as a representation learning problem and the diffe rent network architectures we tried\, how we suited it to our needs\, what worked/didn't work and the challenges we faced along the way.\n\nTopics C overed:\n\n- How we reframed the problem\n- Different network architecture s we tried and their results.\n- Examples of success cases which had faile d previously.\n- Infrastructure and scaling challenges\n\n### Infrastructu re Challenges [~ 5 mins]\n\nSolving the representation problem didn't nece ssarily solve the similarity search problem. We only had a way to sufficie ntly represent all the product information on the vector space. This secti on will cover the infrastructure challenges\, the options we considered an d how we ended up choosing FAISS.\n\nTopics Covered:\n\n- Challenges\, Con straints\n- Re-evaluating Elasticsearch\n- Evaluating FAISS\n- Key bencmar ks\n\n### Conlusion [~ 2 mins]\n\n---\n\n### Speaker bio\n\nAbishek is a m ember of the data science team at Semantics3\, which offers data and AI so lutions for ecommerce marketplaces (catalog generation & enrichment\, sell er on-boarding) and logistics companies (HTS/tariff classification\, attri bute enrichment). Among these\, Abishek is the lead data scientist working on product matching and catalog generation.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115015Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/similarity-search-for- product-matching-semantics3-Qx9qKgrCf9d3N5qNJjjzMt BEGIN:VALARM ACTION:display DESCRIPTION:Similarity search for product matching at Semantics3 in Audito rium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Improving product discovery via hierarchical recommendations DTSTART:20190726T044000Z DTEND:20190726T051000Z DTSTAMP:20260421T132223Z UID:session/GeuutwnCBBNXqLcC3vJQWn@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Lecture,Full talk of 40 mins,Scheduled rehearsal,H as potential. Mentor proposer. CREATED:20190711T062128Z DESCRIPTION:- Introduction to recommendation system at Flipkart\n- Problem in hand\n- Our journey towards recommending collections\n- How hierarchic al product taxonomies can be leveraged to solve cold-start problem and imp roving product discovery\n- Relevance algorithm@scale\n- Captivating findi ngs and results\n\n### Speaker bio\n\nNeha is a software developer with Re commendation team at Flipkart. She has worked on building scalable systems for product recommendations and personalisation. In the past she has also worked on Natural Language Processing. She is interested in building robu st data processing pipelines at scale\, and applying Machine Learning to s olve challenging problems . She has graduated from IIT BHU. While not work ing on official projects\, she involves herself in technical writing and b logging. She also contributes to the open source world by answering techni cal questions.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T115004Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/improving-product-disc overy-via-hierarchical-recommendations-GeuutwnCBBNXqLcC3vJQWn BEGIN:VALARM ACTION:display DESCRIPTION:Improving product discovery via hierarchical recommendations i n Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:BoF on ML and Kubernetes DTSTART:20190726T044500Z DTEND:20190726T052500Z DTSTAMP:20260421T132223Z UID:session/YEgjhab2amRt7PwevS7jdL@hasgeek.com SEQUENCE:2 CREATED:20190724T162407Z DESCRIPTION:This BoF discusses the benefits of using Kubernetes as the und erlying infrastructure for ML.\n\nWe will discuss the following key aspect s of why it is a natural fit for Kubernetes to be any ML platform’s infr astructure:
\n1. Understanding different stakeholders of an ML platform - skill sets required\, specific roles and access levels etc\n2. Different workflows in ML and how it is backed by infrastructure - how compatibilit y of Kubernetes simplifies this\n3. Portability of Kubernetes makes ML bot h cloud native\, on-premise and on the edge\n4. Scalability and distribute d computing are a built in part of Kubernetes - deployments are easy as we ll.\n5. Downsides of Kubernetes - adoption within teams\n6. Short discussi on with current users of Kubernetes for ML\n\n### Speaker bio\n\nKrishna D urai\, Ravishanker KS\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:BoF on ML and Kubernetes in Birds of Feather (BOF) area in 5 m inutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Demystifying Social Network Analysis (SNA) DTSTART:20190726T044500Z DTEND:20190726T052500Z DTSTAMP:20260421T132223Z UID:session/5Bo98U7jnKM6wFffXYjsZt@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk of 40 mins,Full talk of 40 mins CREATED:20190708T051854Z DESCRIPTION:1. Subject introduction and motivation\n2. Key concepts and te rminology\n3. Network Measures\n4. Tools\, software used\n5. Analytical te chniques\n6. Applications\n• Indian elections 2014\n• #MeToo Movement\ n• Indian elections 2019\n• Legal eagles on Twitter\n\n### Speaker bio \n\nI am a researcher and data scientist. My research interests are health care\, e-commerce and social media. There are exciting possibilities\, int eresting insights waiting to be uncovered by network analysis. The talk is wide in its coverage and applications to ensure everyone\, from geeks to politicians or doctors or journalists are kept onboard and learn something new as much as add to others.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114956Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/demystifying-social-ne twork-analysis-sna-5Bo98U7jnKM6wFffXYjsZt BEGIN:VALARM ACTION:display DESCRIPTION:Demystifying Social Network Analysis (SNA) in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190726T051000Z DTEND:20190726T054000Z DTSTAMP:20260421T132223Z UID:session/6CXqcToS7XNWbTjMQG4owb@hasgeek.com SEQUENCE:0 CREATED:20190717T052318Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T052325Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190726T052500Z DTEND:20190726T055500Z DTSTAMP:20260421T132223Z UID:session/WJzjjhERGWqQJQytgiaWqv@hasgeek.com SEQUENCE:0 CREATED:20190723T102854Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190723T102854Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning beverage break DTSTART:20190726T052500Z DTEND:20190726T055500Z DTSTAMP:20260421T132223Z UID:session/MGAaVNvyuheG4wDg6HDZvU@hasgeek.com SEQUENCE:0 CREATED:20190515T063251Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T052406Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning beverage break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Tutorial: Meet TransmogrifAI\, Open Source AutoML powering Salesfo rce Einstein DTSTART:20190726T054000Z DTEND:20190726T074000Z DTSTAMP:20260421T132223Z UID:session/5mnrWWzhPvriLpso8dymBr@hasgeek.com SEQUENCE:2 CATEGORIES:Intermediate,Very company specific: maybe,No slides and preview video ,This proposal is NOT a reject yet.,Tutorial CREATED:20190529T061027Z DESCRIPTION:- Introduction\n- Need of Multicloud and multi tenant models\n - Lessons learned while building Einstein platform\n- How traditional mach ine learning works\n- Introducing TransmogrifAI\n- Type Hierarchy\n- Autom atic Feature Engineering across text\, categorical\, numerical\, spatial f eatures\n- Handling label leakage\n- Autmatic Model Selection and hyper pa rameter tuning\n- Models supported currently\n- Demo\n- Uses cases being s olved in production\n- Summary\n\n### Speaker bio\n\nRajdeep is leading In dustries Einstein team at Salesforce which is leveraging TransmogirfAI bas ed data pipeline to solve ML problems across domains. He has overall 19 ye ars of Software experience and has written 3 books in area on ML and DL.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/meet-transmogrifai-ope n-source-automl-powering-salesforce-einstein-5mnrWWzhPvriLpso8dymBr BEGIN:VALARM ACTION:display DESCRIPTION:Tutorial: Meet TransmogrifAI\, Open Source AutoML powering Sal esforce Einstein in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Challenges and approaches for instrumenting and cleaning 'real'/ ugly data DTSTART:20190726T055500Z DTEND:20190726T065500Z DTSTAMP:20260421T132223Z UID:session/H7zzV2GoB9GDjdpcBmEeTn@hasgeek.com SEQUENCE:2 CATEGORIES:Birds of a Feather session of 1 hour CREATED:20190722T084348Z DESCRIPTION:In this session we will share our experience on multiple thing s associated with collecting/instrumenting data and converting it to usefu l/accessible form. Specifically\, the focus will be on emerging data in sp ace of agriculture\, biology\, telemetry\, images. Few of the topics which would be discussed are:\n\n1. Dealing with incomplete data\n2. Data accur acy issues \n3. Tools for data dictionary \n4. Process of data cleaning \n 5. QA for right data instrumentation\n\n##We will cover the following topi cs in this session: \n\n- Data basics like meta data store\, data dictiona ry\, documentation of data flow -- challenges and tooling.\n- Types of dat a issues including discovery of known unknowns: incomplete data\, improper instrumentation\, corrupted pipeline\, varying quality\, quality of label s -- what are the reasons?\n- Categorising sources and types of issues -- ways of dealing with each type.\n- Steps/milestones in journey for better data: legacy data\, new data\n- How to know what to instrument -- various approaches.\n- Suggestions and recommendations to approach the problem.\n\ n### Speaker bio\n\nKranthi Mitra -- Principal Data Scientist at Swiggy\nR aghotham Sripadraj -- Senior Data Scientist at Ericsson\nElvis Joel D'Souz a -- Director of Product Engineering at Sensara\nKarnam Vasudeva Rao -- Se nior Scientist in the data science team at Bayer\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/challenges-and-approac hes-for-instrumenting-and-cleaning-real-ugly-data-H7zzV2GoB9GDjdpcBmEeTn BEGIN:VALARM ACTION:display DESCRIPTION:Challenges and approaches for instrumenting and cleaning 'rea l'/ ugly data in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Technology to counter misinformation/disinformation DTSTART:20190726T055500Z DTEND:20190726T063500Z DTSTAMP:20260421T132223Z UID:session/PssP7E4MZaHY2edm8ph8Ba@hasgeek.com SEQUENCE:3 CATEGORIES:Crisp talk,Intermediate CREATED:20190530T110502Z DESCRIPTION:This is a call for the tech community to come together and cre ate open source technlogy which can help fight misinformation/disinformati on. The technology created will be useful not only in the Indian context b ut also in the global context.\n\n### Speaker bio\n\nI'm the co-founder of Alt News\, a fact-checking website which has been working on the issue of misinformation/disinformation since Feb 2017. Previously\, I have a backg round in software engineering having worked in the wireless/embedded field for over 13 years in India\, US and Vietnam. With my experience with Alt News and my previous experience of working on some cutting edge technlogie s\, I am able to envision how technology can play a very important role in this fight against misinformation/disinformation.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114947Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/technology-to-counter- misinformation-disinformation-PssP7E4MZaHY2edm8ph8Ba BEGIN:VALARM ACTION:display DESCRIPTION:Technology to counter misinformation/disinformation in Auditor ium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of a Feather: ML in production DTSTART:20190726T061500Z DTEND:20190726T070000Z DTSTAMP:20260421T132223Z UID:session/ABywEtF8LvmUedKaGoJGJp@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190722T061722Z DESCRIPTION:TBA\n\n### Speaker bio\n\nTBA\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/birds-of-a-feather-ml- in-production-ABywEtF8LvmUedKaGoJGJp BEGIN:VALARM ACTION:display DESCRIPTION:Birds of a Feather: ML in production in Birds of Feather (BOF) area in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Sponsored talk: Feed generation at ShareChat DTSTART:20190726T063500Z DTEND:20190726T071000Z DTSTAMP:20260421T132223Z UID:session/2DbSKUfJmbySJMpNvzUy9S@hasgeek.com SEQUENCE:3 CATEGORIES:Short talk of 20 mins,Short talk of 20 mins CREATED:20190627T083455Z DESCRIPTION:ShareChat is India’s largest vernacular social network platf orm built to enable next generation of India’s internet users. ShareChat is available in 14 vernacular languages. At ShareChat our data is fresh\, with most users coming online for first time\, our primary goal is to ser ver most relevant content to the users at appropriate time. In this talk w e will discuss the new challenges these first time internet user present. We will motivate the feed generation problem and give a walkthrough of Fee d Generation algorithm at ShareChat.\n\n1. Introduction to ShareChat\n2. R ecommendation Systems Landscape: Evolution of recommender systems from Gro up Lens to Netflix and advent of Collaborative Filtering.\n3. Deep Learnin g in Recommender Systems: Deep Learning Algorithms in academia and industr y which try to solve recommendation problem at scale.\n4. Feed Generation Problem: What is feed generation problem and how it is different from clas sic recommendation systems.\n5. Data Challenges: Challenges in designing f eed generation for ShareChat and unique insights that ShareChat’s data p resents.\n6. ShareChat’s approach to solving Feed Generation\n7. Other p roblems at ShareChat\n\n### Speaker bio\n\nAyush is currently a lead data scientist at ShareChat. He designs algorithms for content-relevance and fe ed-generation. Ayush comes with past experience of working on a varied set of data science problems in different domains including Healthcare\, Fin- Tech\, Life Sciences and Manufacturing domains.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114938Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/feed-generation-sharec hat-2DbSKUfJmbySJMpNvzUy9S BEGIN:VALARM ACTION:display DESCRIPTION:Sponsored talk: Feed generation at ShareChat in Auditorium 1 i n 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: Controlling narratives on Twitter DTSTART:20190726T065500Z DTEND:20190726T075500Z DTSTAMP:20260421T132223Z UID:session/Gg5awvWovhSay944sNuWPb@hasgeek.com SEQUENCE:1 CREATED:20190717T080759Z DESCRIPTION:This BOF will go into detail on how narratives are controlled on Twitter via: \n\n1. Fake news\n2. Bots\n3. Paid followers\n4. Trend man ipulation \n5. Trolls\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230108T103046Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: Controlling narratives on Twit ter in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On data science and its applicatio ns in agriculture DTSTART:20190726T070000Z DTEND:20190726T074500Z DTSTAMP:20260421T132223Z UID:session/16MPxYdsfmuFhfagqXy8Bj@hasgeek.com SEQUENCE:0 CREATED:20190717T103030Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T084608Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On data science and its applic ations in agriculture in Birds of Feather (BOF) area in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Data security and startups: make the ends meet DTSTART:20190726T071000Z DTEND:20190726T073500Z DTSTAMP:20260421T132223Z UID:session/Cz8KZXYsHzFmSyEKJuajfx@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Lecture,Short talk of 20 mins CREATED:20190712T163319Z DESCRIPTION:I will cover the following in my talk:\n\n - Data security and how it differs from application security/penetration testing\n - Ground r ealities of data security\n - Data security and how to implement it withou t compromising the organization’s growth\n - Why is data security needed when I have perimeter security\, firewalls\, intrusion detection system\, etc in place?\n - What are you protecting when you are enforcing data sec urity?\n - Technical solutions for implementing data security\, and why th is approach is better than instituting processes for protecting data?\n - The big picture: GDPR compliance once data security is implemented?\n - Se curity standards and compliance requirements for launching in countries wh ere GDPR exists.\n - How to structure/re-balance tech to be GDPR-ready.\n - High state of data security and GDPR and it’s relationship with micros ervices.\n - Metrics to track and evaluating how your company is doing on data security parameters.\n\n### Speaker bio\n\nShadab has led Black Ops t eams err.. Information Security teams as a specialist with unicorns like O la\, Flipkart and large scale Internet firms like Adobe. An engineer by he art with out of the box thinking.\nHe has good hands-on experience in E-co mmerce\, payment gateways\, mobile security\, logistic product\, Digital s igning\, Container/Infra Security\, plugging security as part of SDLC to n ame and few others.\nHe has bootstrapped security engineering team multipl e times from scratch. He has experience around building security automatio n\, building real-time detection of attack anomalies\, evangelizing securi ty\, compliance\, cryptography and making sure the product security is kep t the tallest. \n\nCurrently\, he heads Information security\, Privacy and Trust @Hotstar\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114929Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/a-data-security-framew ork-which-you-can-implement-in-your-company-Cz8KZXYsHzFmSyEKJuajfx BEGIN:VALARM ACTION:display DESCRIPTION:Data security and startups: make the ends meet in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190726T073500Z DTEND:20190726T083000Z DTSTAMP:20260421T132223Z UID:session/NDHiPWSPH1AZJyHHYdbFLt@hasgeek.com SEQUENCE:0 CREATED:20190717T080929Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T080948Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190726T074000Z DTEND:20190726T083000Z DTSTAMP:20260421T132223Z UID:session/7FUaQgAdr7pjRQaLpDevJC@hasgeek.com SEQUENCE:0 CREATED:20190515T063828Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T114312Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch break DTSTART:20190726T075500Z DTEND:20190726T083000Z DTSTAMP:20260421T132223Z UID:session/B8xB53McC35bUvqM9w3cT1@hasgeek.com SEQUENCE:0 CREATED:20190717T080821Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190722T083842Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How GO-FOOD built a query semantics engine to help you find food f aster DTSTART:20190726T083000Z DTEND:20190726T091000Z DTSTAMP:20260421T132223Z UID:session/Rr7j2a6UCKAFuk4UWKYyAb@hasgeek.com SEQUENCE:2 CATEGORIES:Beginner,Full talk of 40 mins,Submit revised slides within 7 da ys CREATED:20190703T085835Z DESCRIPTION:There are multiple components we built that can be grouped und er the umbrella of Query Understanding. In this talk\, I will briefly cove r the following:\n\n* Spell Correction\n* Intent Classification\n* Query E xpansion\n* Knowledge Graphs\n* Autosuggest and Autocomplete\n\nTwo of the most important components of the Query Understanding Workflow are Intent Classification and Query Expansion: this talk will cover both of these in further detail and will go over the following topics with respect to the m odels we built:\n\n* Finding the right data to train the models\n* Choosin g the right algorithm: Word2Vec versus Doc2Vec\n* Available open-source li braries and implementations\n* Building the end-to-end pipeline for model training and deployment\n* Experimenting and Iterating for continuous impr ovement\n\n### Speaker bio\n\nIshita has been working as a Data Scientist since 2016 with product-based startups in understanding business concerns in various domains and formulating them as technical problems that can be solved using data and ML. Her current work at GO-JEK involves end-to-end d evelopment of ML projects\, by working as part of a product team in defini ng\, prototyping and implementing data science models within the product. She has also published a book on "Applied Supervised Learning with Python" with publisher Packt.\n\nIshita has completed her Masters’ degree in Hi gh Performance Computing with Data Science from the University of Edinburg h\, UK and her Bachelors’ degree with Honours in Physics from St. Stephe n’s College\, Delhi.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/how-go-food-built-a-qu ery-semantics-engine-to-help-you-find-food-faster-Rr7j2a6UCKAFuk4UWKYyAb BEGIN:VALARM ACTION:display DESCRIPTION:How GO-FOOD built a query semantics engine to help you find fo od faster in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On Interpretability of ML Models DTSTART:20190726T083000Z DTEND:20190726T093000Z DTSTAMP:20260421T132223Z UID:session/56dp5ake18tGCRg9EUooQa@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190717T073625Z DESCRIPTION:- Why is model interpretability important?\n- Types of model i ntrpretability.\n- Popular libraries which explain interpretability.\n- Tr ade off between interpretability and accuracy.\n\n### Speaker bio\n\n##Fac ilitators:\n\n- Namrata Hanspal\n- Ramprakash R\n- Fathat Habib\n- Aditya Patel\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/bof-on-interpretabilit y-of-ml-models-56dp5ake18tGCRg9EUooQa BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On Interpretability of ML Mode ls in Birds of Feather (BOF) area in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On multi-tenancy in ML -- the SaaS perspective DTSTART:20190726T083000Z DTEND:20190726T093500Z DTSTAMP:20260421T132223Z UID:session/oAKxLWgMKnsRAKQRRDu5v@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190719T092205Z DESCRIPTION:Multi-tenant ML models - enabling self-serve of ML capabilitie s at scale\nCustomizable ML models - enabling customers to seamlessly tail or models to fit their needs\nVertical-specific models - training models a t an industry-level\, making ML features available to new customers from d ay 0\nLearning from continuous feedback at scale\nCase studies - Automatio n through Virtual assistants (chatbots\, intent detection\, lead scoring)\ n\n### Speaker bio\n\nSwaminathan Padmanabhan is the Director for Datascie nce initiatives at Freshworks. He has been with Freshworks Inc since 2017 and has built a team of ~20 ML Engineers and Data scientists\, who are ba sed out of both the Bangalore and Chennai R&D centers. Prior to joining F reshworks\, Swaminathan was heading the Datascience function at Olacabs\, Bangalore\; where he was leading a team of ~35 Research Engineers\, System s engineers and Data scientists. An alumnus of IIT Madras\, he was also a ssociated in the past with Inmobi and Yahoo Inc.\n\nWe'll also have 1-2 ex ternal participants and 2-3 other participants from the Datascience team a t Freshworks join the BoF session.\n\nOther Freshworks participants:\nSuvr at Hiran - Lead ML engineer \nVarun Nathan - Lead Data scientist \n\nExter nal participants: TBD\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/multi-tenancy-in-machi ne-learning-the-saas-perspective-oAKxLWgMKnsRAKQRRDu5v BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On multi-tenancy in ML -- the SaaS perspective in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Why data privacy is critical for robust data management? DTSTART:20190726T083000Z DTEND:20190726T085000Z DTSTAMP:20260421T132223Z UID:session/Jygsg3A8KpWBDNgt54UK57@hasgeek.com SEQUENCE:3 CATEGORIES:Full talk of 40 mins CREATED:20190627T082241Z DESCRIPTION:* Data science is not just a job\n * operationalization vs. exploration\n * empiricism\n * democratization\, "citizen" data sci ence\n * The role and future of open source\n * Two types of OSS\n * Software isn't just code\n * Crowdsourcing innovation\n * What comes n ext?\n * Hardware innovation\n * Desegregating computing\n * The coming age of inference engines\n\n### Speaker bio\n\nPeter Wang is a co-f ounder of Anaconda\, Inc.\, where he is CTO and leads the Open Source and Community Innovation team. He has been developing commercial scientific co mputing and visualization software for 20 years. He has extensive experien ce in software design and development across a broad range of areas\, incl uding 3D graphics\, geophysics\, large data simulation and visualization\, financial risk modeling\, and medical imaging.\n\nAs a creator of the PyD ata community and conferences\, he devotes time and energy to growing the Python data science community and advocating and teaching Python at confer ences around the world. Peter holds a BA in Physics from Cornell Universit y.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114917Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/state-of-data-science- machine-learning-Jygsg3A8KpWBDNgt54UK57 BEGIN:VALARM ACTION:display DESCRIPTION:Why data privacy is critical for robust data management? in Au ditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How to build blazingly fast distributed computing like Apache Spar k In-house? DTSTART:20190726T085000Z DTEND:20190726T091500Z DTSTAMP:20260421T132223Z UID:session/RhqYWY9q8kd2xr4GKv92rd@hasgeek.com SEQUENCE:3 CATEGORIES:Advanced,Short talk of 20 mins,Scheduled rehearsal CREATED:20190529T061215Z DESCRIPTION:In this talk we will take care of below questions and explain the same followed by a demo is the system build.\nWhat is business motivat ion to build Spark like(or better) distributed processing framework in-hou se?\nWhy distributed frameworks like Spark will not work for us in long ru n and why we need something else?\nWhat are basic design layers\, data str uctures and algorithms required to build one such a system?\nWhat are the benchmark results and how it works better than Spark for us?\nDemo run of the framework.\n\n### Speaker bio\n\nSpeaker: Upendra Singh: Full Stack Da ta Scientist\, 11 years of experience in distributed algorithm development \, distributed computing and ML\nContributor: Lallit Parsai: Data Engineer -2\, 6 years experience in Data Engineering and Distributed Systems.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114902Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/how-to-build-blazingly -fast-distributed-computing-like-apache-spark-in-house-RhqYWY9q8kd2xr4GKv9 2rd BEGIN:VALARM ACTION:display DESCRIPTION:How to build blazingly fast distributed computing like Apache Spark In-house? in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Demo session: Samantar: an open assistive translation framework fo r Indic Language DTSTART:20190726T091000Z DTEND:20190726T093000Z DTSTAMP:20260421T132223Z UID:session/VtzeWtiuyghtWk5wtQDqir@hasgeek.com SEQUENCE:2 CATEGORIES:Short talk of 20 mins CREATED:20190718T091550Z DESCRIPTION:In this case study\, we will be discussing the following areas \n\n- Requirement of a translation systems in social context\n- State of c urrent translation systems\n- Collation of open corpora for various langua ges\n- Parallel corpora collection in various sectors like Budgets\, Judic iary etc..\n- Various approaches of translation systems\n- Natural Languag e Processing techniques crucial to translation systems.\n- Evaluation and usage of existing open source translation systems like moses\, open NMT et c..\n- Highlevel architecture of samantar\n- Various ways of interacting a nd colloboration with the framework\n- Domain adoptation with the translat ion framework\n- Road ahead\n\nThis session addresses following points/are as\n\n- Overview of NLP for Indic Languages\n- Open translation systems an d their applications\n- Open Parallel Corpora available\n- A Indic languag e translation framework\n- Challenges working with Indic Languages for NLP \n- Domain based translation mechanisms\n\n### Speaker bio\n\nDeepthi Chan d has been on the forefront of the data-for-good movement in India. Over t he last six years he has dabbled in various roles from an application deve loper in MNCs to a data strategist for various civil-society organizations and government agencies. He is co-founder and director of CivicDataLab\, where he works to harness data\, tech\, design and social science to stren gthen civic-engagements in India. He has been leading DataKind Bangalore\, a community of data scientists volunteering their time to help non-profit s do data-driven decision making over the weekends. He is determined to wo rk on key issues in social sector using open-source software\, open data a nd algorithmic research.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/an-open-assistive-tran slation-framework-for-indic-language-samantar-VtzeWtiuyghtWk5wtQDqir BEGIN:VALARM ACTION:display DESCRIPTION:Demo session: Samantar: an open assistive translation framewor k for Indic Language in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:A journey through Cosmos to understand users DTSTART:20190726T091500Z DTEND:20190726T094500Z DTSTAMP:20260421T132223Z UID:session/Cweivs4BXbfbDG8dndo78f@hasgeek.com SEQUENCE:3 CATEGORIES:Intermediate,Full talk of 40 mins,Confirmed,Strong accept CREATED:20190530T061741Z DESCRIPTION:The topics we will be covering in this talk: \n1. Introduction - Briefly provide business context to appreciate the need to solve this p roblem\, and challenges involved.\n2. The factors driving the decision to choose Cosmos DB as our backend store.\n3\, Key insights into what drives cost of the store\, and various gotchas involved when designing such a sys tem. \n4. How to optimize the cost and bring intelligence to enable auto-s calability.\n5. The need for building a multi version concurrency control and how to achieve it to enable parallel writes with multiple schema versi ons for the same record.\n6. The tradeoff between readability and storage cost\, and how to get the best of both worlds by building an avro library to enable inflight abbreviated compression.\n\n### Speaker bio\n\nAvinash Ramakanth: \nTech lead at Inmobi\, MSc Computer Systems Indian Institute o f Science. \nI was part of the group which experimented and conceptualized the design for building the user inference systems for Inmobi DSP. My pri or experience for the past 4 years\, involve understanding user data at In mobi and building large scale systems to provide inferences for enabling i ntelligent ad serving. This work spans across building large scale stream processing systems\, ML pipelines to make inferences and various big data applications.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20240119T114848Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/a-journey-through-cosm os-to-understand-users-Cweivs4BXbfbDG8dndo78f BEGIN:VALARM ACTION:display DESCRIPTION:A journey through Cosmos to understand users in Auditorium 1 i n 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Demo sessions: share your idea\; get feedback DTSTART:20190726T093000Z DTEND:20190726T104500Z DTSTAMP:20260421T132223Z UID:session/TJsJ4sGXq6J4RyTdjoExjH@hasgeek.com SEQUENCE:0 CREATED:20190717T053238Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190718T121111Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Demo sessions: share your idea\; get feedback in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: Anomaly detection at large scale ( for data security) DTSTART:20190726T093000Z DTEND:20190726T103000Z DTSTAMP:20260421T132223Z UID:session/AL4GAetUbRUyu4f2nXL1MU@hasgeek.com SEQUENCE:2 CREATED:20190717T102611Z DESCRIPTION:**This BOF is a conversation on the following questions: \n\n1 . Performing security analytics on logs for different attacks at a scale o f more than 10 TBB logs a day.\n2. Managing tools with high dependency for security like OpenSoc/Apache Metron.\n3. Justifying infra cost for securi ty analytics where ROI is a question.\n4. What is bare minimum we can do t o stop real-time attacks without WAF (i.e.\, too many false positives).\n5 . Experiences on\, and better ways of\, building anomaly detection tool an d how to handle rules threshold.\n6. Classify training data for ML models. \n7. Break glass appraoch or process setup that is followed for doing ano maly detection at large scale?\n\n**Who should participate: participants w ho have first-hand experience in doing anomaly detection for data security at large scale. This BOF will be in small size\, since the goal is to sha re first-hand experiences with this challenge\, at large scale.\n\n### Spe aker bio\n\nShadab has led Black Ops teams err.. Information Security team s as a specialist with unicorns like Ola\, Flipkart and large scale Intern et firms like Adobe. An engineer by heart with out of the box thinking.\nH e has good hands-on experience in E-commerce\, payment gateways\, mobile s ecurity\, logistic product\, Digital signing\, Container/Infra Security\, plugging security as part of SDLC to name and few others.\nHe has bootstra pped security engineering team multiple times from scratch. He has experie nce around building security automation\, building real-time detection of attack anomalies\, evangelizing security\, compliance\, cryptography and m aking sure the product security is kept the tallest.\n\nCurrently\, he hea ds Information security\, Privacy and Trust @Hotstar\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: Anomaly detection at large sca le (for data security) in Birds of Feather (BOF) area in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On intent classification and perso nalization DTSTART:20190726T094000Z DTEND:20190726T104500Z DTSTAMP:20260421T132223Z UID:session/CApCzXjtceHWGaZbBsKhEs@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190717T072809Z DESCRIPTION:- Types of problems in which intent classification is essentia l vs. good-to-have\n(examples - chatbots\, knowledge extraction from texts \, ...)\n - Related to the previous point\, given certain use-cases\, how specific/generic do the "intent categories" have to be?\n - Chatbots\, w hich are probably the most visible applications utilizing intent classific ation and personalization can be discussed in a little bit more detail her e.\n - If treating personalization as a standalone goal by itself\, what "features" and "techniques" are worth looking at?\n - Given that personal perceptions and attitudes change quite frequently\, what sort of historic al bias vs. recency bias is required for these systems. On this note\, can background knowledge bases be used to better adapt the models on smaller initial training datasets.\n\n##Key takeaways\n\nA general overview of the field of intent classification and personalization as a bare minimum. Hop efully\, we can also learn about which techniques to consider when startin g off\, when to use them\, and of-course\, when not to use them. Common st ories with expectation-outcome mismatches would also be useful takeaways.\ n\n### Speaker bio\n\n##Facilitators:\n\n- Aditya Patel\n- Ishita Mathur\n - Ramanan Balakrishnan\n- Maulik Soneji\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/intent-classification- and-personalization-CApCzXjtceHWGaZbBsKhEs BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On intent classification and p ersonalization in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Flash talk on MUDPIPE: malicious URL detection for phishing identi fication and prevention DTSTART:20190726T094500Z DTEND:20190726T095000Z DTSTAMP:20260421T132223Z UID:session/Qtn11f64bbZmgmoMdXeREo@hasgeek.com SEQUENCE:2 CATEGORIES:Short talk of 20 mins CREATED:20190710T065030Z DESCRIPTION:1. Introduction to Phishing & Social Engineering\n2. Threat ac tors and vectors in phishing exploitation attacks\n3. Determinants at play while evaluating website genuineness\n4. How to build your own Machine Le arning Model for phishing detection\n5. Demo of an existing model and mode l evaluation\n6. Factors to be considered while deploying the model in pro duction\n\n### Speaker bio\n\nArjun is a security professional with divers e experience in architecting\, designing\, implementing & supporting IT Se curity & Vulnerability Management solutions in Enterprise & Cloud environm ents. He is an information security enthusiast with diverse experience in areas like Application Security\, Security Architecture\, DevSecOps\, Clou d Security & Machine Learning. Currently\, Arjun is currently working as a Security Architect ensuring end-to-end implementation\, design and govern ance of security measures an e-commerce platform\, aimed at brand protecti on and improving customer confidence. He is currently developing products that aid in phishing detection for the enterprise and ensure that defenses are in place to counter this threat.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/mudpipe-malicious-url- detection-for-phishing-identification-and-prevention-Qtn11f64bbZmgmoMdXeRE o BEGIN:VALARM ACTION:display DESCRIPTION:Flash talk on MUDPIPE: malicious URL detection for phishing id entification and prevention in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Flash talks by the audience DTSTART:20190726T095000Z DTEND:20190726T101500Z DTSTAMP:20260421T132223Z UID:session/QgWAZGmLwJGv4GEsscwzqU@hasgeek.com SEQUENCE:0 CREATED:20190710T074119Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190717T124259Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Flash talks by the audience in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190726T101500Z DTEND:20190726T104500Z DTSTAMP:20260421T132223Z UID:session/LLjVzZEyKnc84akwL2GbC8@hasgeek.com SEQUENCE:0 CREATED:20190515T063911Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190718T121201Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Unpacking the Learning Paradigms DTSTART:20190726T103000Z DTEND:20190726T111500Z DTSTAMP:20260421T132223Z UID:session/F12ZdFNtYxjLZBio8FcJQM@hasgeek.com SEQUENCE:2 CATEGORIES:Birds of a Feather session of 1 hour CREATED:20190724T054646Z DESCRIPTION:It starts all simple. You predict the price in Boston housing data and understand that this is **Supervised Learning**. You reduce the d imensions in the Iris flower data and understand that this is **Unsupervis ed Learning**. Then you want to know how the Chess & Go data was used and understand there is **Reinforcement Learning**. \n\nYou move to time-serie s data and now you have something like **Auto-Regressive learning**. Or da bble in text data\, and try to get your head around all these word vectors and language models. Soon you are reading about an alphabet soup of *suff ix-paradigm-Learning* – **Semi-Supervised**\, **Self-Supervised**\, **We ak-Supervised** ... and now you are struggling to make sense of it all. Th row in a bit of statistical model literature: **Generative Learning** vs. **Discriminative Learning**\, **Frequentist Learning** vs. **Bayesian Lear ning** and it no longer looks simple anymore.\n\nThis BoF session is to ha ve an open dialogue on the **Learning Paradigms** and start to unpack them to build a better mental model around them. Some of questions we would be keen to unpack are:\n\n1. Why is it important to have a mental model arou nd the learning paradigms? \n2. How to think about learning from data as a spectrum of algorithmic (and model) techniques\, rather than neat categor isation of learning buckets?\n3. What are some useful analogies and constr ucts that can help both describe and explain these in a way that builds intuition and cognition?\n4. How to successfully navigate the surge of new techniques\, models and tricks that keep emerging in ML papers and librar ies?\n5. How can we develop a better vocabulary for Machine Learning to be better able to explain to the business and general audience what is reall y happening?\n\n### Speaker bio\n\n- [Chris Stucchio](https://www.chrisstu cchio.com)\n- [Amit Kapoor](https://amitkaps.com)\n- ...\n- ...\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Birds of Feather (BOF) area - NIMHANS Convention Centre\nBengalur u\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/unpacking-the-learning -paradigms-F12ZdFNtYxjLZBio8FcJQM BEGIN:VALARM ACTION:display DESCRIPTION:Unpacking the Learning Paradigms in Birds of Feather (BOF) are a in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190726T104500Z DTEND:20190726T111500Z DTSTAMP:20260421T132223Z UID:session/78gTfs323KSK5K5C1Ryxg3@hasgeek.com SEQUENCE:0 CREATED:20190720T083418Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190720T083420Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How we built a ML model to predict proteins for insecticidal acti vity? DTSTART:20190726T104500Z DTEND:20190726T112500Z DTSTAMP:20260421T132223Z UID:session/Ey9rbnQ1j861msiKUpvboD@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk of 40 mins,Short talk of 20 mins CREATED:20190712T082105Z DESCRIPTION:1. What are insecticidal proteins?\n2. Why machine learning fo r protein activity identification?\n3. Different approaches used by resear chers\n4. Why not traditional methods?\n5. iFeature - a Python tool kit\n5 a. Why did we choose iFeature?\n5b. What features iFeature has?\n5c. How w e adopted it for our need?\n5d. What were the challenges?\n5e. How did we overcome those?\n6. Key learnings\n\n### Speaker bio\n\nDr. Karnam Vasudev a Rao is presently working as Senior Scientist-Data Science\, at Monsanto (Subsidiary of Bayer)\, Bengaluru\, India since 2009. Prior to this Vasu h as pursued his PhD from Max-Planck Institute For Biochemistry\, Munich\, G ermany. He has enormous experience working in research organizations in In dia and abroad. He is involved in developing data science products in his organization and in mentoring budding Data Scientists within and outside t he organization.\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/how-we-built-a-ml-mode l-to-predict-proteins-for-insecticidal-activity-Ey9rbnQ1j861msiKUpvboD BEGIN:VALARM ACTION:display DESCRIPTION:How we built a ML model to predict proteins for insecticidal activity? in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening beverage break DTSTART:20190726T104500Z DTEND:20190726T111500Z DTSTAMP:20260421T132223Z UID:session/EnFKNhxt51qNTKRbYYPTEa@hasgeek.com SEQUENCE:0 CREATED:20190717T053534Z DESCRIPTION:\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20190718T121134Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening beverage break in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Birds of Feather (BOF) session: On ML Model Management DTSTART:20190726T111500Z DTEND:20190726T120000Z DTSTAMP:20260421T132223Z UID:session/DseaPALyW2AgeswgEikYDA@hasgeek.com SEQUENCE:2 CATEGORIES:BOF session of 1 hour,Birds of a Feather session of 1 hour CREATED:20190720T083349Z DESCRIPTION:- Managing ML models in production is non-trivial. What are th e challenges and concerns of machine learning management lifecycle?\n- Wha t is machine learning model management?\n\n### Speaker bio\n\n**Ravi Ranja n** is working as Senior Data Scientist at Publicis Sapient. He is part of Centre of Excellence and responsible for building machine learning model at scale. He has worked on multiple engagements with clients mainly from A utomobile\, Banking\, Retail and Insurance industry across geographies. In current role\, he is working on Hyper-personalized recommendation system for Automobile industry focused on Machine Learning\, Deep learning\, Real time data processing on large scale data using MLflow and Kubeflow. \nHe h olds Bachelor degree in Computer Science with proficiency course in Reinfo rcement Learning from IISc\, Bangalore.\n\n- Krishna Durai\n- Ravishankar Babu\n GEO:12.9431582;77.5964488824009 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 3 - NIMHANS Convention Centre\nBengaluru\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2019/schedule/bof-ml-model-managemen t-DseaPALyW2AgeswgEikYDA BEGIN:VALARM ACTION:display DESCRIPTION:Birds of Feather (BOF) session: On ML Model Management in Audi torium 3 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR