BEGIN:VCALENDAR VERSION:2.0 PRODID:-//HasGeek//NONSGML Funnel//EN DESCRIPTION:A conference on big data and analytics X-WR-CALDESC:A conference on big data and analytics NAME:The Fifth Elephant 2014 X-WR-CALNAME:The Fifth Elephant 2014 REFRESH-INTERVAL;VALUE=DURATION:PT12H SUMMARY:The Fifth Elephant 2014 TIMEZONE-ID:Asia/Kolkata X-PUBLISHED-TTL:PT12H X-WR-TIMEZONE:Asia/Kolkata BEGIN:VEVENT SUMMARY:Check-in DTSTART:20140723T040000Z DTEND:20140723T043000Z DTSTAMP:20260421T123212Z UID:session/41sXfBL1gH1MGWQpydvtzG@hasgeek.com SEQUENCE:0 CREATED:20140527T025158Z DESCRIPTION:\n LAST-MODIFIED:20140527T054718Z LOCATION:Auditorium - TERI\, Domlur\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Machine Learning using R : Crash course in Classification Methods DTSTART:20140723T043000Z DTEND:20140723T073000Z DTSTAMP:20260421T123212Z UID:session/QvLc1FqscSazCoHocAEeWP@hasgeek.com SEQUENCE:2 CATEGORIES:Workshops,Beginner CREATED:20140623T124249Z DESCRIPTION:The following topics would be covered. The format would be a b it of theory and then implementation using R\n\nIntroduction to Machine le arning\n1) Types of Learning (Supervised/Unsupervised/Reinforced)\n2) Intr oduction to Generalization\n3) Train/Test/Validation Datasets\n4) Bias – Variance tradeoff\n5) Overfitting\n6) Cross-validation\n7) Regularization \n8) Grid Search\n9) Hyperparameter Optimization\n10) Feature Selection/Tr ansformation\na. Greedy feature selection (forward\, backward\, st epwise)\nb. Non-linear transformations\, Kernels \n\n\nClassificat ion Techniques covered:\n1) Linear Regression\n2) Logistic Regression\n3) LASSO\, Ridge and Elastic net regression\n4) kNN\n5) Discriminant Analysis \n6) Decision Trees\, CART\, CHAID\n7) Support Vector Machines\n8) Naïve Bayes\n9) Ensemble Methods\na. Boosting\nb. Bagging\nc. Random Forest\nd. Regularized Random Forest\ne. Gradient Boosting Machines\n\n\nUnsupervised learning techniques covered:\n1) Dimensionality Reduction: Principal Comp onent Analysis\n2) K-Means clustering\n\nIllustrating common pitfalls\n1) Data snooping\n2) Occam’s Razor\n\nBig Data Analytics (*need AWS credit for implementation.)\n1) Introduction to Big Data and Hadoop\n2) R and Big Data\na. Hadoop\nb. Linear Model\nc. Random Forest\n\n### Speaker bio\n\n Data Analytics professional at Cisco Systems India Pvt Ltd.\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - TERI\, Domlur\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/machine-learning-using -r-crash-course-in-classification-methods-QvLc1FqscSazCoHocAEeWP BEGIN:VALARM ACTION:display DESCRIPTION:Machine Learning using R : Crash course in Classification Met hods in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20140723T073000Z DTEND:20140723T083000Z DTSTAMP:20260421T123212Z UID:session/BfzHg9yLtKUXnpvMwLPfCS@hasgeek.com SEQUENCE:0 CREATED:20140527T025251Z DESCRIPTION:\n LAST-MODIFIED:20140527T025254Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Building distributed search applications using Apache SOLR DTSTART:20140723T083000Z DTEND:20140723T113000Z DTSTAMP:20260421T123212Z UID:session/4xr7gMXN72Da6BYn9JZfy4@hasgeek.com SEQUENCE:2 CATEGORIES:Workshops,Beginner CREATED:20140617T095810Z DESCRIPTION:For the workshop\, we will index and search data from 'StackEx change' sites using dumps available [here](https://archive.org/details/sta ckexchange) and build backend for following demo application: [saumitra.me /solrdemo/](http://saumitra.me/solrdemo/). (Tested only on Chrome and Fire fox).\n\nAgenda:\n\n1. What is Solr? Use cases and architecture\n2. Solr s chema\, config\, tokenizers and filters\n3. Indexing data:\n * From dis k using SolrJ\n * Importing from database(MySQL) with DataImport Handle r\n4. Querying Solr (Filters\, Faceting\, highlighting\, sorting\, groupin g\, boosting\, range\, function and fuzzy queries)\n5. Using 'More Like Th is' component to show similar docs\n6. Adding 'Auto Suggest' component to auto complete user queries\n7. Using 'Clustering' component to cluster sim ilar results.\n8. SolrCloud\n * Architecture\n * Setting up a multin ode cluster with Zookeeper\n * Creating a distributed index\n * Coll ections API\n9. Solr Admin UI\n10. Understanding Solr performance factors\ n11. Solr vs. ElasticSearch\n\n### Speaker bio\n\n* Engineer at [Glassbeam ](http://glassbeam.com) working on machine data search and analytics\n* Or ganizer of [Bangalore-Baby-Apache-Solr](http://www.meetup.com/Bangalore-Ba by-Apache-Solr-Group/) meetup group\, where we organize sessions to help p eople get started with Solr\n* [https://twitter.com/_saumitra_](https://tw itter.com/_saumitra_)\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - TERI\, Domlur\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/building-distributed-s earch-applications-using-apache-solr-4xr7gMXN72Da6BYn9JZfy4 BEGIN:VALARM ACTION:display DESCRIPTION:Building distributed search applications using Apache SOLR in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Check-in DTSTART:20140724T041500Z DTEND:20140724T043000Z DTSTAMP:20260421T123212Z UID:session/XTGEmnqSpvypDuUiMsZ7UF@hasgeek.com SEQUENCE:0 CREATED:20140527T025345Z DESCRIPTION:\n LAST-MODIFIED:20140527T025348Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Real world machine learning DTSTART:20140724T043000Z DTEND:20140724T073000Z DTSTAMP:20260421T123212Z UID:session/E5Ph8Qcbdmn8mMwYXityNN@hasgeek.com SEQUENCE:2 CATEGORIES:Workshops,Intermediate CREATED:20140721T050718Z DESCRIPTION:Machine learning has evolved to a very popular\, rapidly chang ing\, sometimes over-hyped domain with extremely diverse set of ideas. Dis cussion about machine learning often tends to get lost into jargon of tool s\, market buzzwords\, libraries and diverted from real purpose which is i nsights! This workshop will focus on insights and practical applications.\ n\n### What background is essential to understand this workshop ?\n\nThis is introductory session and participants are not expected to know 100 mach ine learning algorithms or have a PhD in Maths. That said\, the following are bare minimum requirements\,\n\n* Reasonable knowledge of at-least some programming language (C\, C++\, Python\, Ruby\, Java\, R\, Matlab\, Julia and so on). \n\n* Some familiarity with machine learning (for example\, i t should be enough to vaguely understand what linear regression is)\n\n* C uriosity!\n\n[Those who are not familiar with Python are advised to go thr ough the tutorial given here](https://docs.python.org/2/tutorial). Python is an extremely elegant and simple language\, you'd get started in no time !\n\n### Speaker bio\n\nHarshad is senior data scientist at Sokrati\, a di gital advertising startup based out of Pune\, where he works closely with the engineering team to extract meaning out of millions of data points fro m advertising world. He's been applying machine learning to real world pro blems in telecom\, banking and advertising since last 4 years. He has most ly worked with tools like R\, Python\, SAS and lately fallen in love with Clojure ecosystem too. He conducted a similar session in Fifth Elephant 20 13\, focussed on R and text mining. Harshad holds a master's degree in Ope rations Research from IIT\, Mumbai\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - TERI\, Domlur\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/real-world-machine-lea rning-E5Ph8Qcbdmn8mMwYXityNN BEGIN:VALARM ACTION:display DESCRIPTION:Real world machine learning in Auditorium in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20140724T073000Z DTEND:20140724T083000Z DTSTAMP:20260421T123212Z UID:session/L1bQDYpPx6RCQsWCMhHVUB@hasgeek.com SEQUENCE:0 CREATED:20140527T025418Z DESCRIPTION:\n LAST-MODIFIED:20140527T025421Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Getting your hands dirty with Aerospike DTSTART:20140724T083000Z DTEND:20140724T113000Z DTSTAMP:20260421T123212Z UID:session/2DHysqhBrKdNssKizDY7Ph@hasgeek.com SEQUENCE:2 CATEGORIES:Sponsored workshop,Intermediate CREATED:20140702T032014Z DESCRIPTION:We will go over the following in the workshop\n1. Talk about t he architecture of Aerospike\n2. Install\, setup\, configure\n3. Use tools built around aerospike\n4. Coding using its API (java/node.js)\n5. Using Aerospike in existing app frameworks (TBD)\n6. Using Aerospike on Amazon c loud.\n7. Unleash the high performance of Aerospike\n\nWe will organize th e people into teams to encourage discussion and mutual help. Also\, in the later part of the workshop\, the team will use their individual EC2 machi nes to form a cluster and learn how to deal with cluster.\n\n### Speaker b io\n\nSunil Sayyaparaju\, Engineering Lead at Aerospike\, has over 9 years experience working on different types of SQL RDBMS solutions\, such as si ngle machine (monolithic)\, in-memory\, distributed shared-disk\, and dist ributed shared-nothing architectures with emphasis in transaction manageme nt\, storage\, access\, performance tuning\, and recovery areas.\n\nSunil currently leads Aerospike's Bangalore office\, working on their distribute d shared-nothing NoSQL solution. Aerospike is a high-performance\, self-ba lancing\, immediately consistent\, distributed NoSQL database. Aerospike a lso has an add on product for replication across data-centers over WAN whi ch supports different complex topologies.\n\n* we will have multiple prese nters\n LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium - TERI\, Domlur\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/getting-your-hands-dir ty-with-aerospike-2DHysqhBrKdNssKizDY7Ph BEGIN:VALARM ACTION:display DESCRIPTION:Getting your hands dirty with Aerospike in Auditorium in 5 min utes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Check-in and breakfast DTSTART:20140725T030000Z DTEND:20140725T034500Z DTSTAMP:20260421T123212Z UID:session/Xw8no16jGrxpAw4cD4aatS@hasgeek.com SEQUENCE:0 CREATED:20140527T025500Z DESCRIPTION:\n LAST-MODIFIED:20140630T165817Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in and breakfast in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Introductions to HasGeek\, The Fifth Elephant 2014\, EP DTSTART:20140725T034500Z DTEND:20140725T040000Z DTSTAMP:20260421T123212Z UID:session/KwcELTLZgGnnHtEU5Zr5qQ@hasgeek.com SEQUENCE:0 CREATED:20140527T025528Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140630T165953Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Introductions to HasGeek\, The Fifth Elephant 2014\, EP in Aud itorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:'Know Your Customer!' - Advanced Data Science for Audience Segment ation DTSTART:20140725T040000Z DTEND:20140725T044500Z DTSTAMP:20260421T123212Z UID:session/KY1U9RqALnVCMfD9enNYyB@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Advanced CREATED:20140704T081723Z DESCRIPTION:Audience Segmentation is a very important practical necessity in pretty much every field. Whether it is internet subscribers or paytv su bscribers there is an intense need from the advertisers and service provid ers to know who is living in a household and what their demography profile s are and what their interests are?\n\nAn ensemble of techniques which inc lude advanced linear and non-linear dimensionality reduction\, unsupervise d learning algorithms\, bayesian predictive analytics and expert systems c ome together to form a compelling pragmatic data science solution stack fo r the big data environment.\n\n### Speaker bio\n\nI am Prabhakar Srinivasa n. I work as a data scientist at Cisco. I have invented a unique technique to do customer segmentation which works for the PayTv domain of Cisco pro ducts running on Big data infrastructure. I wish to share this success sto ry with the community with the hope that others who are trying to solve a similar problem can gain some practical insight on doing data science on b ig data stack that really gives business value.\n\nI have successfully dep loyed my invention in Europe for some Telecom giants. I have presented thi s technique during talks in Monaco during Cisco's internal conferences. It is time to share this success story with the wider Big data community.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/know-your-customer--ad vanced-data-science-for-audience-segmentation-KY1U9RqALnVCMfD9enNYyB BEGIN:VALARM ACTION:display DESCRIPTION:'Know Your Customer!' - Advanced Data Science for Audience Seg mentation in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Unified analytics platform for Bigdata DTSTART:20140725T040000Z DTEND:20140725T044500Z DTSTAMP:20260421T123212Z UID:session/NQcBUreyeVpP2oKdNLxo8g@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T110528Z DESCRIPTION:Conventional columnar databases (RDBMS) systems lend themselve s well for interactive SQL queries over reasonably small datasets in the o rder of 10-100s of GB\, while hadoop based warehouses operate well over la rge datasets in the order of TBs and PBs and scales fairly linearly. Thoug h there have been some improvements recently in storage structures in the Hadoop warehouses such as ORC\, queries over hadoop still typically adopts a full scan approach. Choosing between these different data stores based on cost of storage\, concurrency\, scalability and performance is fairly c omplex and not easy for most users. This talk presents Grill\, the new ana lytics platform for InMobi\, a system built at InMobi to precisely solve t his problem on top Hive metastore. \n\nThe Hive metastore in its current s tate allows users to represent structured data in simple tables. However\, it does not allow expressing relationships or richer DWH concepts like fa cts\, dimensions and etc. With Hive data cubes\, users can query data stor ed in HDFS\, S3\, Redshift etc\, with a single query language and schema. Underlying execution engines like Hive\, Impala\, Shark etc can be plugged in and utilized at run time. The execution engine used is transparent to the user. The system provides a unified logical schema to users consisting of cubes\, facts and dimensions\; and users can issue queries at a concep tual level without knowing about roll-up intervals\, partitions\, data typ es\, underlying storage and table relationships\; they will be figured out automatically.\n\n### Speaker bio\n\nAmareshwari is currently working as Architect in platform team at Inmobi\, where she works on Hadoop and relat ed projects for data collection and analytics. She is member of Apache Had oop PMC and is Apache Hive committer. She has been working on Hadoop and i ts eco system since 2007. Prior to Inmobi\, she was working with Yahoo! in core Hadoop team. She holds bachelor's degree in computer science and eng ineering from National institute of technology\, Waragal\, India\; and mas ter's degree in Internet science and engineering from Indian Institute of Science (IISc)\, Bangalore\, India.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/unified-analytics-plat form-for-bigdata-NQcBUreyeVpP2oKdNLxo8g BEGIN:VALARM ACTION:display DESCRIPTION:Unified analytics platform for Bigdata in Auditorium 1 in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How to build a Data Stack from scratch DTSTART:20140725T044500Z DTEND:20140725T053000Z DTSTAMP:20260421T123212Z UID:session/V2AjYGfQfnUtN1LZhLBKWL@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T111158Z DESCRIPTION:In the talk I will talk about my experience of how to build a data stack from stratch. I have built a big data analytics stack at Akamai and Inmobi before and am currently building one now at Helpshift. These a re three different domains - Content Delivery Networks (Akamai)\, Mobile A dvertising (Inmobi) and now Customer Service (Helpshift).\n\nMore specific ally\, my talk will try to cover these questions and more\n\n* What are th e different components of an analytics stack and what function does each l ayer have ?\n* How do you choose the right software for different layers o f your analytics data stack ?\n* Do you use real-time analytics or batch p rocessing is right for you ? What are the costs/benefits of both ?\n* What is the relation between statistical and probabilistic techniques ? Which to choose when ?\n* How to decide on the right structure and storage for y our data and how they influence your analytics stack ?\n* How to decide on the right metrics for your business and how they influence your analytics stack ?\n\nI will use specific industry examples how each of these questi ons were answered differently in different contexts. I will also talk the factors that influenced these decisions and how they influenced the final output and architecture.\n\n### Speaker bio\n\nVinayak is an early adopter of technologies having worked across diverse and complex computer systems including embedded systems\, networking\, large-scale distributed systems and data-processing systems. He has more than a decade of experience in h ardcore product development & software/deployment architecture.\n\nHe has led engineering teams at Akamai\, Inmobi and Helpshift to build big data s tacks from scratch. He organised one of the first Cloudcamps and Barcamps in India. He co-founded Headstart\, a grass-roots community driven by volu nteers for helping startups. Other than his interests in tech and startups \, he is an avid traveller and amateur photographer.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/how-to-build-a-data-st ack-from-scratch-V2AjYGfQfnUtN1LZhLBKWL BEGIN:VALARM ACTION:display DESCRIPTION:How to build a Data Stack from scratch in Auditorium 1 in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Data sciences (is) in fashion @ Myntra DTSTART:20140725T044500Z DTEND:20140725T053000Z DTSTAMP:20260421T123212Z UID:session/EH9Bf9MvUtq9MESMancEft@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T081650Z DESCRIPTION:Fashion is a hard category to sell online given most of the pu rchases happen on impulse. It fundamentally differs from selling categorie s like mobiles which are more driven through reviews and ratings. Data sci ences help bring a differentiating angle to selling fashion and can be app lied in a variety of fashion e-tailing problems ranging from product ranki ngs (the extreme form of which is personalised store for every user)\, sto re organisation and navigation\, better customer engagement\, better offer creation\, better merchandising decisions\, etc. We @ Myntra are working towards these specific problems and would love to talk about the approache s that have worked for us.\n\nWe are personalising the customer experience in multiple ways\n- by personalising customer communications through vari ous channels\n- by personalising the website to tailor to customers' prefe rences \n- by creating unique customer specific offers\n\nWe would as well be talking about the data platform which powers all these efforts.\n\n### Speaker bio\n\nDivya Alok\, Devashish\, Debdoot\nData scientists working at myntra. We look at tons of data and actively look towards creating data products used by our customers - external and internal.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/data-sciences-is-in-fa shion-myntra-EH9Bf9MvUtq9MESMancEft BEGIN:VALARM ACTION:display DESCRIPTION:Data sciences (is) in fashion @ Myntra in Auditorium 2 in 5 mi nutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning tea break DTSTART:20140725T053000Z DTEND:20140725T060000Z DTSTAMP:20260421T123212Z UID:session/UFEFoZbszDfYYQSup71puR@hasgeek.com SEQUENCE:0 CREATED:20140527T025658Z DESCRIPTION:\n LAST-MODIFIED:20140529T032006Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Sponsored session: What enterprises can learn from real-time biddi ng DTSTART:20140725T060000Z DTEND:20140725T064500Z DTSTAMP:20260421T123212Z UID:session/XisbAqoWV8Hpy6kjV3dEEs@hasgeek.com SEQUENCE:0 CREATED:20140527T025635Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20200619T062516Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Sponsored session: What enterprises can learn from real-time b idding in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Serving user intent : Facebook style notifications using HBase and Event streams DTSTART:20140725T064500Z DTEND:20140725T073000Z DTSTAMP:20260421T123212Z UID:session/6E5cBiTfjoM6VZmGjJGJUd@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T114013Z DESCRIPTION:Relevant and Personalized notifications in near real-time is a great way of serving user intent. The intent may vary - say liking a Face book update as compared to a price drop for a browsed product on an e-comm erce website. The system characteristics and solution patterns in both the se instances may be very similar though.\n\nThis talk will cover the desig n of the [Flipkart](http://www.flipkart.com) Notifications platform. The t echniques and technologies used to serve product related intent can be eas ily applied to a different domain. This talk will also introduce projects that were Open Sourced while building the platform.\n\nArchitecture\, Desi gn patterns and technologies used in this system include:\n\n* Pre-creatin g data that matches user intent - so as to significantly reduce data servi ng latencies\n* Storing immutable events and interpreting change\n* Event driven architectures(EDA) and its variant Staged EDA (SEDA) using technolo gies like RabbitMQ and Mule.\n* Complex Event Processing (CEP) using techn ologies like Esper\n* Data stores like HBase that organize data between me mory and disk as Log Structured Merged (LSM) trees - leveraging Disk trans fer better over Disk seek\n* A data serving API that is resilient to failu res and latencies - using Hystrix and Netty\n\nThe talk uses a typical e-c ommerce experience where user intent is either implicit or interpreted fro m actions - for example a user browsing a product of interest\, adding an item to a shopping cart or adding it for future reference via a wish-list. In a dynamic e-commerce marketplace\, product data (such as price\, stock quantity) is constantly changing across millions of listed products even as user intent is being expressed on the website. User intent may be seen as one Event stream while Product attribute changes is another. An interse ction of these two streams is the Notification data. An efficient data sto re that can store and serve tens of millions of such notifications with ve ry low latencies is the Notification service.\n\nThe following projects we re open sourced before or when building the Notifications Platform : \n\n* [Trooper](https://github.com/regunathb/Trooper)\n* [Phantom](https://gith ub.com/Flipkart/phantom)\n\nThe talk will also feature a live view of the data serving metrics with millisecond response times.\n\n### Speaker bio\n \nArchitect and Open source committer. My areas of interest are Distribute d Systems\, Big Data\, Text Mining and Data Stores.\n\nMy experience as Ar chitect includes:\n\n* Building the World's largest biometric identity pla tform in [Aadhaar](http://uidai.gov.in/)\n* Customer facing Mobile and Web platforms at India's leading e-Commerce company - [Flipkart](http://www.f lipkart.com)\n\nMost of my work in recent years has been around OSS - usin g it to build large scale systems and in contributing projects back to the community. Some of my OSS work is downloaded and used worldwide:\n\n* [Ph antom](https://github.com/Flipkart/Phantom)\n* [Trooper](https://github.co m/regunathb/Trooper)\n* [Insight](http://sourceforge.net/projects/mindtree insight/)\n\nActive projects on github : https://github.com/regunathb\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/serving-user-intent-fa cebook-style-notifications-using-hbase-and-event-streams-6E5cBiTfjoM6VZmGj JGJUd BEGIN:VALARM ACTION:display DESCRIPTION:Serving user intent : Facebook style notifications using HBase and Event streams in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:How we do research on high frequency finance DTSTART:20140725T064500Z DTEND:20140725T073000Z DTSTAMP:20260421T123212Z UID:session/UGEarTipc6Ycs4x7tNhHq@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T113532Z DESCRIPTION:It is the story of the Finance Research Group at Indira Gandhi \nInstitute of Development Research. How a set of researchers from non\nco mputer science backgrounds helped build a computing facility which\nis one of the largest amongst finance labs in India.\n\nI will be talking about our challenges and solutions we came up with:\n\n** Problem statement\n\n- Analyse Indian financial markets data to assess market quality after intr oduction of algorithmic trading.\n- Do computation to judge high frequency trading activity in the Indian markets.\n- Manage and validate gigabytes of data inflow every day.\n- No metadata for existing data.\n- Utilise exi sting infrastructure.\n- Prepare for disasters.\n- Keep it all simple: the y're not geeks!\n- Automate!\n- Low budget (!!)\n\n** Managing new giganti c data\n\n - From the National Stock Exchange\, Mumbai.\n - \\#1 excha nge in the world in terms of volumes traded daily.\n - Around 15GB compr essed data every day.\n - Derived data uncompresses to around 120GB.\n - 170 million orders daily in Futures and Options markets.\n - Flat fil e structure over NFS.\n\n** Deciding the hardware\n\n - Computations req uire loading entire files into memory.\n - Need fast network to transfer this data for computation.\n - Reliable storage\n - Low cost cluster: beowulf\n - A Storage Area Network with in-house Network Attached Stora ge\n\n** Building a new server room\n\n - Heating issues\n - Space con straints\n - Adequate power supply\n\n** Software\n\n - Metadata syste m: home-brewed using R and MySQL\n - Home-grown systems for daily proces sing: downloading\, verification\, checksumming\, and validation.\n - Mo nitoring system: Icinga\n - Batch scheduler: SLURM\n - Logging server: Sentry\n - Data shared over NFS\n - Cluster computation: R + TCP sock ets/MPI\n - Backups on Amazon Glacier\n - Languages used: R\, C\, Pyth on\, Shell\, Java\n - Use C where R gives up\n - Better algorithms\n\n ** Open source contribution\n\n - ifrogs R package\n - eventstudies R package\n * Future\n - LDDB\n\n### Speaker bio\n\nI am a Research Pr ogrammer in the Finance Research Group at the Indira Gandhi Institute of D evelopment Research in Mumbai. My experience includes working with a start up\, a corporate and research groups in Delhi and Mumbai. My areas of inte rest include high performance computing\, high frequency trading\, and fin ancial data management.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/how-we-do-research-on- high-frequency-finance-UGEarTipc6Ycs4x7tNhHq BEGIN:VALARM ACTION:display DESCRIPTION:How we do research on high frequency finance in Auditorium 2 i n 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20140725T073000Z DTEND:20140725T083000Z DTSTAMP:20260421T123212Z UID:session/YGZcVMCdtiTBbGjXPWDXiU@hasgeek.com SEQUENCE:0 CREATED:20140527T025814Z DESCRIPTION:\n LAST-MODIFIED:20140527T025824Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Large Scale Modelling and Analytics Challenges at a Payments Compa ny DTSTART:20140725T083000Z DTEND:20140725T091500Z DTSTAMP:20260421T123212Z UID:session/ScCVnB78jwk1id7hf37NPE@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T113500Z DESCRIPTION:At American Express\, we serve tens of millions of credit card members who transact at several million merchants globally. Apart from th is transaction level data\, there is petabyte scale data generated from c ard members' other interactions with American Express:- through visiting and interacting with our website\, phone and chat interactions with custom er care representatives\, the surveys we conduct to gauge the pulse of our customers\, etc. All these result in large scale data coming from multip le modalities. In this talk\, we will focus on some key challenges which a rise from dealing with this data. The talk will first give a broad overvie w of the various challenges in the context of machine learning in the paym ents industry and then focus in some details on a particular problem of mo delling purchase intent of credit card holders.\n\n### Speaker bio\n\nhttp s://www.linkedin.com/in/subhajitsanyal\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/large-scale-modelling- and-analytics-challenges-at-a-payments-company-ScCVnB78jwk1id7hf37NPE BEGIN:VALARM ACTION:display DESCRIPTION:Large Scale Modelling and Analytics Challenges at a Payments C ompany in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Circuitscape - A Case Study on Scientific Computing DTSTART:20140725T083000Z DTEND:20140725T091500Z DTSTAMP:20260421T123212Z UID:session/LungitQdoGFWLxLtrKEhC9@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate,Lecture CREATED:20140704T114000Z DESCRIPTION:Circuitscape is an open-source program\, which borrows algorit hms from electronic circuit theory to predict patterns of movement\, gene flow\, and genetic differentiation among plant and animal populations in h eterogeneous landscapes. It is used by academics\, policy makers\, and gov ernments around the world in conservation planning.\n\nCreated by Brad McR ae and Viral B. Shah\, it equates life forms to electrons\, the landscape as a grid of resistances and the movement of life forms across a landscape as current flowing through a circuit. Interestingly\, this has been able to model reality much better than many other approaches. Circuitscape has been used to model raster landscapes containing as many as 20 million cell s\, covering vast geographies over thousands of square kilometres\, result ing in jobs that run over days to compute wildlife corridors.\n\nThis talk is about Circuitscape and its application. I will cite the example of Cir cuitscape usage: "Connectivity of Tiger (Panthera tigris) Populations in t he Human-Influenced Forest Mosaic of Central India” during the talk\, an d how this concept can be applied across other domains.\n\n["Connectivity of Tiger (Panthera tigris) Populations in the Human-Influenced Forest Mosa ic of Central India”](http://www.plosone.org/article/info%3Adoi%2F10.137 1%2Fjournal.pone.0077980)\n\n### Speaker bio\n\nI am one of the co-creator s of the Julia programming language and Circuitscape.\nhttp://in.linkedin. com/in/viralbshah\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/circuitscape-a-case-st udy-on-scientific-computing-LungitQdoGFWLxLtrKEhC9 BEGIN:VALARM ACTION:display DESCRIPTION:Circuitscape - A Case Study on Scientific Computing in Auditor ium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Why we built the most adopted Polyglot Object Mapper for NoSQL? DTSTART:20140725T091500Z DTEND:20140725T100000Z DTSTAMP:20260421T123212Z UID:session/5TJkNk4tffhYVbv9rUswFv@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T115220Z DESCRIPTION:NoSQL Datastores are tough. But JPA interface of dealing with Datastores is easy\, well known and has been popular with developers since long. So we combined both the two things and created Kundera. \n\nKundera is innovative in many ways:\n• Kundera supports polyglot persistence i. e. an application can use multiple and any combination of datastores. All the hard work related to mapping\, persisting\, reading\, indexing and tra nsacting object model across multiple datastores is handled by Kundera\,\n • Kundera supports 8 NoSQL datatsores and any RDBMS.\n• Kundera provid es very easy and well known interfaces like JPA and REST to interact with all the datastores. Hence it hides all the complexities of NoSQL stores an d reduces the learning curve enormously.\n• Kundera also facilitates eas y migration of your existing hibernate/JPA powered applications to NoSQL.\ n• Kundera is extremely extensible and one can easily add support for ne w datastores with very minimal effort.\n\nImpact of this product:\n• It has made the task of a NoSQL developer extremely easy. \n• It has simpli fied working with 8 leading NoSQL datastores\; Cassandra\, MongoDB\, HBase \, Neo4j\, Redis\, Oracle NoSQL\, CouchBase and ElasticSearch.\n• It has enabled 10 very big scale applications to use NoSQL and move to productio n.\n• It has enabled number of people in open source community to build exciting solutions (like NoSQL Datastore Migrator\, NoSQL Data Viewer) on top of Kundera.\n\nWhile coming up with such a product that talks to so ma ny different types of datastore\, there have been challenges. And the lear nings. The talk would cover these challenges and learnings.\n\n### Speaker bio\n\nVivek Shrivastava\, Technical Architect & NoSQL Evangelist\, Impet us Technologies \n\nVivek Shrivastava has over 14 years of experi ence on product development\, designing and engineering enterprise grade s olutions for Finance\, Travel\, Retail and Interactive Television domains. His expertise includes NoSQL Data-stores\, Cloud Computing and BigData. V ivek has spearheaded the design & architecture of several large & cloud sc ale solutions for data lifecycle management at petabyte scale. Vivek also leads the development at Open source project Kundera which is a JPA 2.0 co mpliant Object-NoSQL Mapping Library and is a major advocate of Polyglot P ersistence.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/why-we-built-the-most- adopted-polyglot-object-mapper-for-nosql-5TJkNk4tffhYVbv9rUswFv BEGIN:VALARM ACTION:display DESCRIPTION:Why we built the most adopted Polyglot Object Mapper for NoSQL ? in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Scaling Spatial Data - OpenStreetMap as Infrastructure. DTSTART:20140725T091500Z DTEND:20140725T100000Z DTSTAMP:20260421T123212Z UID:session/VTpsPK2QwTrr7nKqV6gEFF@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T115806Z DESCRIPTION:OpenStreetMap has over 9 years of lessons in managing large sc ale crowdsourced spatial data with open source infrastructure. The project now has over 1\,600\,000 contributors out which about 3000 users map ever y day. They create over 16\,35\,608 nodes. The changes are replicated acro ss several official and unoffical instances and all the maps rendered with OpenStreetMap data are updated in under a minute on the Internet.\n\nWith PostgreSQL and PostGIS as the backbone of data storage\, OpenStreetMap ha s [created an ecosystem of tools](http://wiki.openstreetmap.org/wiki/Compo nent_Overview) that is now available for anyone to use for different kinds of spatial data. The data can be processed\, queried and styled in severa l ways. We will discuss setting up the OpenStreetMap infrastructure for cu stom data models and explore the edit-style-render toolchain. There is a r ich API and database dump/restore mechanisms that is built into the infras tructure. OpenStreetMap also takes quality assurance seriously that there is a process in place for identifying and rectifying vandalism or errors.\ n\n### Speaker bio\n\n[Sajjad Anwar](http://twitter.com/geohacker) is a ha cktivist and programmer based in Bangalore. He works in the research and d esign of data infrastructure\, analytics and infographics. Being involved with OpenStreetMap for over 5 years\, he has extensive experience working with spatial data and advocates open geographic data. He helps organisatio ns to build and maintain their data infrastructure. He is found working wi th other technologists\, social activists and researchers in education\, h uman rights and policy making. Along with two others\, he runs the [geohac kers.in](http://geohackers.in) collective.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/scaling-spatial-data-o penstreetmap-as-infrastructure-VTpsPK2QwTrr7nKqV6gEFF BEGIN:VALARM ACTION:display DESCRIPTION:Scaling Spatial Data - OpenStreetMap as Infrastructure. in Aud itorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:High-tea – sponsored by Flipkart DTSTART:20140725T100000Z DTEND:20140725T104500Z DTSTAMP:20260421T123212Z UID:session/55fN7Pqiby5KWqqdrtHVbe@hasgeek.com SEQUENCE:0 CREATED:20140612T000056Z DESCRIPTION:\n LAST-MODIFIED:20140721T050920Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:High-tea – sponsored by Flipkart in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Using Cascalog and Clojure to make the elephant move! DTSTART:20140725T104500Z DTEND:20140725T110000Z DTSTAMP:20260421T123212Z UID:session/6X9nev2diTWy8ewkGTBied@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20140705T025546Z DESCRIPTION:Machine learning practitioners often spend 70%-80% of their ti me on cleaning\, transforming and scrubbing data. Quick iterations and abi lity to create large number of features out of raw data are pre-requisites for any serious machine learning activity. There is also a large gap betw een a skill sets and objectives of data scientists and programmers. Data s cientists need to slice and dice through real world _big data_. Implementi ng all the low level plumbing of multiple\, dependent Hadoop jobs is a ser ious impediment to this goal. On the other hand\, companies cannot afford to have _lab_ tools running\, as they have large impedance mismatch with p roduction environments.\nWe have seen Cascalog library and functional abst ractions of Clojure providing a sweet spot. It lets data scientists focus on their job of understanding data without worrying about complicated clas s hierarchies or un-intuitive domain specific languages. On the other hand \, Cascalog and Clojure seamlessly integrate with JVM based stack. The res ult is a fun and productive way to process data at scale!\n\n### Speaker b io\n\nHarshad Saykhedkar is senior data scientist at Sokrati\, a digital a dvertising startup based out of Pune. He has experience of applying machin e learning to problems in advertising\, banking and telecom sectors for pa st four years. He has used multiple tools (R\, Python\, SAS) in this journ ey and lately fallen in love with the Clojure ecosystem\, hacking his way to create data processing tools at Sokrati. Harshad holds a master's degre e in Operations Research from IIT\, Mumbai.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/using-cascalog-and-clo jure-to-make-the-elephant-move-6X9nev2diTWy8ewkGTBied BEGIN:VALARM ACTION:display DESCRIPTION:Using Cascalog and Clojure to make the elephant move! in Audit orium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Analytics on Large Scale\, Unstructured\, Dynamic Data using Lambd a Architecture DTSTART:20140725T104500Z DTEND:20140725T113000Z DTSTAMP:20260421T123212Z UID:session/2Cm7dJhKzTqkVCqKzxVUGN@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140704T115156Z DESCRIPTION:Indix is a product intelligence platform. Our catalog has seve ral million products and billions of price points collected from thousands of e-commerce web sites and is constantly growing. We collect product dat a via crawling product pages from these web sites. Our parsers extract pro duct attributes from these pages which are then run through a series of ma chine learning algorithms to classify and extract deeper product attribute s. This data gets deduped between stores and then matched across stores an d is finally fed into our analytics engine which provides insights to our customers.\n\nOur first attempt at building this system around two years a go was chaotic. We were dealing with e-commerce sites whose pages were uns tructured and were constanly changing. Our parsers and machine learning al gorithms were also improving regularly. All this meant that we had to run our algorithms on the entire data set very often. It was not uncommon for our data refreshes to run for days which meant high latency for product an d price data. In addition to that\, our data systems were not human fault tolerant. We had issues where an incorrect algorithm would get accidentall y deployed to production and corrupt the data we were serving. Since our d ata store was mutable and did not mantain these changes\, it was not easy to fix these corruption issues.\n\nWe realized soon that we had to re-thin k about our data system from ground up. We needed a simpler approach that would scale\, be tolerant to human errors and can evolve with our product. \n\nLambda architecture\, coined by Nathan Marz\, the creator of Storm and Cascalog\, seemed like a step in the right direction for us. \nThe system has been in production for more than a year now\, handling 3X more data t han our older system and most importantly is more robust. \n\nLambda archi tecture\, at its core\, is a set of architecture principles that allows bo th batch and real-time or stream data processing to work together while bu ilding immutability\, recomputation and human fault tolerance into the sys tem.\n\nIt has three layers - batch\, serving and speed.\n\nThe batch laye r is responsible for computing arbitrary views on the master data. Our mas ter data is an immutable store in HDFS and we compute views using a series of Map Reduce jobs using Scalding and Spark. Our batch system runs recom putation every day on our entire data set. \n\nThe serving layer indexes a nd exposes precomputed views to be queried ad-hoc with low latency. We us e HBase\, Solr and our own inhouse inmemory implementation for the serving layer.\n\nThe speed layer deals only with new data and compensates for th e high latency updates of the serving layer by creating realtime views. Ou r real time latency requirements are in few hours and not in seconds\, whi ch allows us to use a micro-batch architecture that is a stripped down ver sion of our batch layer and uses the same technologies.\n\nTo get the fina l result\, the batch and realtime views must be queried and the results me rged together.\n\nTopics I will cover \n\n- Why Lambda Architecture? What problems did it solve for us? \n- Technical Challenges encountered in buil ding the lambda architecture\n - Schema Evolution\n - HDFS Small Fil es Issue\n - Code re-use between batch and real time systems \n- Modeli ng the data pipelines for each layer\n- Open problems\n\n### Speaker bio\n \nRajesh Muppalla is a co-founder and Director of Engineering at Indix\, w here he leads the data platform team that is responsible for collecting\, organizing and structuring all the product related data collected from the web.\n\nHe is passionate about big data\, large scale distributed systems \, continuous delivery and algorithms. He also likes mentoring and coachin g developers in pursuit of building better software.\n\nPrior to Indix\, h e was a technical lead on Go-CD\, an agile and release management product\ , at Thoughtworks. The product has been recently open-sourced. \n\nHe is a gold medallist in Computer Science from Pune University. In his final yea r of graduation\, his team represented India at Asia finals of Microsoft I magine Cup (then called Microsoft .NET campus challenge).\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/analytics-on-large-sca le-unstructured-dynamic-data-using-lambda-architecture-2Cm7dJhKzTqkVCqKzxV UGN BEGIN:VALARM ACTION:display DESCRIPTION:Analytics on Large Scale\, Unstructured\, Dynamic Data using L ambda Architecture in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Storing relationships in large data-sets using Graphs DTSTART:20140725T110000Z DTEND:20140725T113000Z DTSTAMP:20260421T123212Z UID:session/DXy2rEmNjvB1Pzr18m9hSM@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Advanced CREATED:20140705T025640Z DESCRIPTION:This work was motivated to store large amounts of linkeddata i n an ad system and make it available for programmatic/analytics consumptio n. \n\nThis talk outlines our journey which started from researching exist ing graphdb's/processing frameworks\, why they didn't work for us at our s cale and then moving on to build something.\nWe will go in depth to explai n the data-structures used and how we supported the tinker-pop graph API specification( used by all graph databases). We will also touch upon how o ur ad-system unique data model allowed us to come up with a fairly simplis tic technique to shard the entire thing and query over it.\n\nTakeaways fr om this talk -\n\n* what are graphdb's\, when should you choose one.\n* di fferent use-cases require different stores.\n* what it takes to build a gr aph store for allo-centric(alike OLAP) graph traversals.\n\n### Speaker bi o\n\nInder Singh - have been working on solving data related problems at I nmobi(World's largest independent ad-network) for the past ~3 years.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/storing-relationships- in-large-data-sets-using-graphs-DXy2rEmNjvB1Pzr18m9hSM BEGIN:VALARM ACTION:display DESCRIPTION:Storing relationships in large data-sets using Graphs in Audit orium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keynote: Realizing Large-scale Distributed Deep Learning Networks over GraphLab DTSTART:20140725T113000Z DTEND:20140725T121500Z DTSTAMP:20260421T123212Z UID:session/HmYDFDLLXnQCKK9eaqsfpB@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140703T095601Z DESCRIPTION:Large scale distributed deep learning networks are the holy gr ail of the machine learning/AI/data science fields. The applications of de ep learning networks are in image processing\, speech recognition and vide o analytics. We have implemented such a network over GraphLab\, the open s ource graph processing framework. As can be expected\, several extensions were made to GraphLab for implementing such deep learning networks. The ke y extension was the ability to run multiple instances of the distributed G raphLab engine in the same cluster – this allows the training data to be parallelized. Another missing abstraction was that of a layer – which c omprises/aggregates several nodes (vertices of GraphLab). An additional re quirement was mass communication between two layers\, the ability to send a message to all vertices that belong to a layer. We have implemented thes e abstractions in GraphLab and have consequently realized the deep learnin g network over a cluster of nodes. We have used this network for Arrhythmi a detection from ECG images. The talk would also articulate some of the pe rformance studies we have conducted on our distributed deep learning netwo rk.\n\n### Speaker bio\n\nDr. Vijay Srinivas Agneeswaran has a Bachelor's degree in Computer Science & Engineering from SVCE\, Madras University (19 98)\, an MS (By Research) from IIT Madras in 2001 and a PhD from IIT Madra s (2008). He was a post-doctoral research fellow in the LSIR Labs\, Swiss Federal Institute of Technology\, Lausanne (EPFL) for a year. He has spent the last seven years creating intellectual property and building products in the big data area in Oracle\, Cogniizant and Impetus\, where he is now Director\, Big Data Labs. He has built PMML support into Spark/Storm and is building a big data governance product for a role-based fine-grained ac cess control inside of Hadoop YARN. He is a professional member of the ACM and the IEEE for the last 8+ years. He has filed patents with US and Euro pean patent office's (with two issued US patents) and published in leading journals and conferences\, including IEEE transactions. His research inte rests include distributed systems - cloud\, grid\, peer-to-peer computing as well as machine learning for Big-Data and other emerging technologies.\ n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/realizing-large-scale- distributed-deep-learning-networks-over-graphlab-HmYDFDLLXnQCKK9eaqsfpB BEGIN:VALARM ACTION:display DESCRIPTION:Keynote: Realizing Large-scale Distributed Deep Learning Netwo rks over GraphLab in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:CIO Panel DTSTART:20140725T121500Z DTEND:20140725T134500Z DTSTAMP:20260421T123212Z UID:session/QWLoSnZYM1WoUQvsGfE6T3@hasgeek.com SEQUENCE:0 CREATED:20140705T030533Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20200619T062516Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:CIO Panel in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Check-in and breakfast DTSTART:20140726T030000Z DTEND:20140726T034500Z DTSTAMP:20260421T123212Z UID:session/QwiEUPZWzvndb9BiVuhVWr@hasgeek.com SEQUENCE:0 CREATED:20140527T030358Z DESCRIPTION:\n LAST-MODIFIED:20140527T030405Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Check-in and breakfast in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Summary of day 1 DTSTART:20140726T034500Z DTEND:20140726T040000Z DTSTAMP:20260421T123212Z UID:session/Biw9YFt1jw1mGPQmc9EC5A@hasgeek.com SEQUENCE:0 CREATED:20140527T030415Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140527T061240Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Summary of day 1 in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Keynote: Personalized medicine and big data - Anu Acharya DTSTART:20140726T040000Z DTEND:20140726T044500Z DTSTAMP:20260421T123212Z UID:session/Sg8JMtyCtioKcLpVWnxhBh@hasgeek.com SEQUENCE:0 CREATED:20140703T094729Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20200619T062516Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Keynote: Personalized medicine and big data - Anu Acharya in A uditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lessons from Elasticsearch in production DTSTART:20140726T044500Z DTEND:20140726T053000Z DTSTAMP:20260421T123212Z UID:session/XcoVLT7xnPJL55pEzL7ZxR@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140705T025737Z DESCRIPTION:The customer-facing side of Helpshift product is a simple chat feature within the app using the Helpshift mobile SDK. The business-facin g side is a complex agent dashboard that helps the agent in processing as many issues as quickly as possible. We will be focusing on this business-f acing side\, the designs we built on top of Elasticsearch and the problems we faced and how we went about solving them\, a few of them are:\n\n1. Ar chitecture 101 - Importance of separating master-only and data-only nodes\ , etc.\n2. How we index documents for each customer - the flaw in having o ne index per customer\, and possible solutions\n3. Multilingual data - the importance of the phonetic plugin\n4. Complex views - the importance of u nderstanding filters and how to combine them\n5. The importance of benchma rking - how we implemented using percolators for live notifications of new issues\n6. Restarts & Upgrades - the importance of disabling shard alloca tion and clustering\n7. Bulk Indexing - the importance of controlling repl ica count\, etc.\n8. Runtime debugging - the importance of cat APIs\, etc. \n\n### Speaker bio\n\nI work in the backend team at Helpshift.com\, a cus tomer service platform for mobile apps\, our customers include Flipboard\, Supercell\, Flipkart\, and several others.\n\nI have previously worked at Automatic.com (the "UI for your car" company)\, Infibeam.com\, Adobe and Yahoo!.\n\nI have also written a couple of Creative Commons-licensed books - A Byte of Python and A Byte of Vim.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/lessons-from-elasticse arch-in-production-XcoVLT7xnPJL55pEzL7ZxR BEGIN:VALARM ACTION:display DESCRIPTION:Lessons from Elasticsearch in production in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Live analytical dashboards at scale - SQL style DTSTART:20140726T044500Z DTEND:20140726T053000Z DTSTAMP:20260421T123212Z UID:session/KGiBPJjfdPy9tUpWDWTPER@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140703T095400Z DESCRIPTION:Fireball is a stream processing engine at Flipkart. It powers real time analytical dashboards to enable business take time-sensitive dec isions\, at scale. Fireball can process millions of events (with flexible\ , json-like schema) per hour that require:\n* executing custom process (us ually SQL-like) to derive business metrics from the incoming events\n* ove r large number of dimensions (on an average 10 dimensions for each measure )\n* with very low latency and ensuring correctness all the time (enabling time-sensitive decision making)\n\nSo how do you build such a system? How do you store such a large amount of time-series data to ensure roll-ups\, drill-downs on different dimensions? In this talk we'll go over the trans formation of a standard stream processing platform and a CEP library into Fireball.\n\n### Speaker bio\n\nI am Architect @ Flipkart and am part of t he Data Platform effort.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/live-analytical-dashbo ards-at-scale-sql-style-KGiBPJjfdPy9tUpWDWTPER BEGIN:VALARM ACTION:display DESCRIPTION:Live analytical dashboards at scale - SQL style in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Morning tea break DTSTART:20140726T053000Z DTEND:20140726T060000Z DTSTAMP:20260421T123212Z UID:session/BHLSMDeFD88hpdf3HHhJtt@hasgeek.com SEQUENCE:0 CREATED:20140527T030451Z DESCRIPTION:\n LAST-MODIFIED:20140529T032136Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Morning tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Scaling real time visualisations for Elections 2014 DTSTART:20140726T060000Z DTEND:20140726T064500Z DTSTAMP:20260421T123212Z UID:session/Zh77RPVHduD3RNwJXSeAY@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140703T095420Z DESCRIPTION:The CNN-IBN Microsoft Election Analytics Center\, which you ca n see the live visualisations at www.bing.com/elections\, served over 10 m illion requests on election day. \n\nThis includes real-time filtering of the election commission results – based on turnouts\, margins\, computat ions of anti-incumbency factors\, alliance groupings\, etc.\n\nThis talk i s about the engineering from Gramener that went into making the site fast and responsive.\n\n### Speaker bio\n\nAnand is the chief data scientist at Gramener. He explores data stories visually with Python and Javascript.\n \nHe blogs at http://www.s-anand.net/\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/scaling-real-time-visu alisations-for-elections-2014-Zh77RPVHduD3RNwJXSeAY BEGIN:VALARM ACTION:display DESCRIPTION:Scaling real time visualisations for Elections 2014 in Auditor ium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Scaling SolrCloud to a large number of collections DTSTART:20140726T060000Z DTEND:20140726T064500Z DTSTAMP:20260421T123212Z UID:session/Ht9FgEpK1cyo2tL1pTMNhZ@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Advanced CREATED:20140705T025754Z DESCRIPTION:The traditional and typical search use case is the one large s earch collection distributed among many nodes and shared by all users. How ever\, there is a class of applications which need a large number of small or medium collections which can be used\, managed and scaled separately. This talk will cover our effort in helping a client set up a large scale S olrCloud setup with thousands of collections running on hundreds of nodes. I will describe the bottlenecks that we found in SolrCloud when running a large number of collections. I will also take you through the multiple fe atures and optimisations that we contributed to Apache Solr to reduce or r emove the choke points in the system. Finally\, I will talk about the benc hmarking process and the lessons learned from the exercise.\n\n### Speaker bio\n\nI am a committer on Apache Lucene/Solr since 2008 as well as a mem ber of the Lucene/Solr project management committee. I've worked at AOL fo r five years on vertical search\, content mangement systems\, social/commu nity platforms and anti-spam systems as well as AOL WebMail's Inbox Search system which uses a highly customized version of Apache Solr to service t ens of millions of users and more than a billion index/search operations a day. I currently works at LucidWorks Inc. on Apache Solr and LucidWorks S earch mostly on the SolrCloud side of things. I also help organize the Ban galore Apache Solr/Lucene Meetup Group which has 350+ members and holds re gular meetings of people interested in Lucene\, Solr and search in general .\n\nhttps://twitter.com/shalinmangar\nhttp://www.meetup.com/Bangalore-Apa che-Solr-Lucene-Group/\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/scaling-solrcloud-to-a -large-number-of-collections-Ht9FgEpK1cyo2tL1pTMNhZ BEGIN:VALARM ACTION:display DESCRIPTION:Scaling SolrCloud to a large number of collections in Auditori um 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:The state of Julia - a fast language for technical computing DTSTART:20140726T064500Z DTEND:20140726T070000Z DTSTAMP:20260421T123212Z UID:session/5qAHzZFv1QqepNWGrJzJdW@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20140705T030042Z DESCRIPTION:I will give a talk introducing Julia for those who have not he ard about it. I will talk about progress of the language since the last ye ar\, and some glimpses into where we are headed. I will also discuss the g rowth of the Julia community.\n\nJulia is a high-level\, high-performance dynamic programming language for technical computing\, with syntax that is familiar to users of other technical computing environments. It provides a sophisticated compiler\, distributed parallel execution\, numerical accu racy\, and an extensive mathematical function library. The library\, large ly written in Julia itself\, also integrates mature\, best-of-breed C and Fortran libraries for linear algebra\, random number generation\, signal p rocessing\, and string processing. In addition\, the Julia developer commu nity is contributing a number of external packages through Julia’s built -in package manager at a rapid pace. IJulia\, a collaboration between the IPython and Julia communities\, provides a powerful browser-based graphica l notebook interface to Julia.\n\nJulia programs are organized around mult iple dispatch\; by defining functions and overloading them for different c ombinations of argument types\, which can also be user-defined. For a more in-depth discussion of the rationale and advantages of Julia over other s ystems\, see the following highlights or read the introduction in the onli ne manual.\n\nhttp://www.julialang.org/\nhttp://docs.julialang.org/\n\n### Speaker bio\n\nI am one of the co-creators of the Julia programming langu age.\n\nhttp://in.linkedin.com/in/viralbshah\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/the-state-of-julia-a-f ast-language-for-technical-computing-5qAHzZFv1QqepNWGrJzJdW BEGIN:VALARM ACTION:display DESCRIPTION:The state of Julia - a fast language for technical computing i n Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Dr. Hadoop – Diagnose your Hadoop Jobs DTSTART:20140726T064500Z DTEND:20140726T070000Z DTSTAMP:20260421T123212Z UID:session/Eu9TzPSSyBnK6ZFfb8GGxo@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Intermediate CREATED:20140705T030019Z DESCRIPTION:This talk is about a tool that we have developed within intuit – Dr. hadoop\, which analyzes your job\, identifies the areas of improv ements and gives recommendations to improve its performance. It collects a ll the history logs\, counters and configuration of your job\, applies a s et of rules and provides recommendations with suggested values and severit y.\n\n### Speaker bio\n\nI am a hadoop performance engineer@Intuit. I have been working on hadoop performance for more than 3 years.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/dr-hadoop-diagnose-you r-hadoop-jobs-Eu9TzPSSyBnK6ZFfb8GGxo BEGIN:VALARM ACTION:display DESCRIPTION:Dr. Hadoop – Diagnose your Hadoop Jobs in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:De-dup on Hadoop DTSTART:20140726T070000Z DTEND:20140726T073000Z DTSTAMP:20260421T123212Z UID:session/JR4pHCJJfoux6JbSkkLzWH@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Beginner CREATED:20140705T030003Z DESCRIPTION:In many enterprises it's commonly seen that business data has a lot of client\, customer\, vendor or product lists in different formats and systems\, many of which are near duplicates.MDM solutions on RDBMS hav e been prominent for many years in almost every enterprise to support mast er data management by removing duplicates\, standardizing data and incorpo rating rules to eliminate incorrect data from entering the system in order to create an authoritative source of master data. MDM on Big data platfor ms like Hadoop have benefits as well as it's own set of challenges when co mpared with the RDBMS counterparts. I will cover them in detail primarily focusing on building this solution on Hadoop.\n\n### Speaker bio\n\nI am D ata Architect at Intuit with 13+ years of experience in BI and Data Analyt ics. Prior to Intuit\, I have worked at Intel\, Oracle and EMC applying BI in Manufacturing\, Finance and Storage Analytics domain.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/de-dup-on-hadoop-JR4pH CJJfoux6JbSkkLzWH BEGIN:VALARM ACTION:display DESCRIPTION:De-dup on Hadoop in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Using Data for Art DTSTART:20140726T070000Z DTEND:20140726T073000Z DTSTAMP:20260421T123212Z UID:session/CnBvtcD5PenztzbfKengZZ@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Beginner CREATED:20140705T030055Z DESCRIPTION:A talk on Data Art for a tech-focused audience must introduce the essence of art (vs design) and the purpose of creating art. I'll intro duce the field with a few teasers (conventional art vs data art) and then introduce the modes in which data can be mapped in the final art form (mos tly visual attributes\, and a bit of other forms of media too). This will be followed by an introduction to Generative Systems in art\, where an art form evolves from data (and isn't completely controlled by an artist) and a categorization of these systems (with hand picked examples) that I had explored recently.\n\n### Speaker bio\n\nI'm Rasagy Sharma\, and I visuali ze data at Microsoft IDC. I'm a recent post graduate from National Institu te of Design\, Bangalore where I pursued Information & Interface Design\, which had a heavy focus on Data Visualization.\n\nI've recently been inter ested in Generative Systems to create art\, and this talk will leverage on my research for my Colloquium paper on Creative Coding\, in which I categ orized ~100 generative art projects and the popular media used.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/using-data-for-art-CnB vtcD5PenztzbfKengZZ BEGIN:VALARM ACTION:display DESCRIPTION:Using Data for Art in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Lunch DTSTART:20140726T073000Z DTEND:20140726T083000Z DTSTAMP:20260421T123212Z UID:session/ipgsvK7pcyY49zaci1wXe@hasgeek.com SEQUENCE:0 CREATED:20140527T031015Z DESCRIPTION:\n LAST-MODIFIED:20140704T082535Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Lunch in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Apache Tez: Accelerating Hadoop Data Pipelines DTSTART:20140726T083000Z DTEND:20140726T091500Z DTSTAMP:20260421T123212Z UID:session/5ALMUy1sYQXsoHkm3qgHCa@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Beginner CREATED:20140705T030129Z DESCRIPTION:Apache Tez is a modern data processing engine designed for YAR N on Hadoop 2. Tez aims to provide high performance and efficiency out of the box\, across the spectrum of low latency queries and heavy-weight batc h processing. With a clear separation between the logical app layer and th e physical data movement layer\, Tez is designed from the ground up to be a platform on top of which a variety of domain specific applications can b e built. Tez has pluggable control and data planes that allow users to plu g in custom data transfer technologies\, concurrency-control and schedulin g policies to meet their exact requirements. The talk will elaborate on th ese features via real use cases from early adopters like Hive\, Pig and Ca scading.\n\n### Speaker bio\n\nGopal works on performance problems in hado op ecosystem. He's involved with the Stinger effort from Hortonworks to im prove the SQL data access layers in Hadoop. He is a contributor to the Apa che Hive project and a committer for the Apache Tez project.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/apache-tez-acceleratin g-hadoop-data-pipelines-5ALMUy1sYQXsoHkm3qgHCa BEGIN:VALARM ACTION:display DESCRIPTION:Apache Tez: Accelerating Hadoop Data Pipelines in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Crafting Visual Stories with Data DTSTART:20140726T083000Z DTEND:20140726T091500Z DTSTAMP:20260421T123212Z UID:session/UgTNaWHpRRoyg8Sjtt42xr@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Beginner CREATED:20140705T030115Z DESCRIPTION:*"I think people have begun to forget how powerful human stori es are\, ex-changing their sense of empathy for a fetishistic fascination with data\, networks\, patterns\, and total information... Really\, the da ta is just part of the story. The human stuff is the main stuff\, and the data should enrich it." - Jonathan Harris*\n\nStories have been recognized for their power of communication & persuasion for centuries. There is an increasingly realisation that we need to operate at this intersection of d ata\, visual and stories to fully harness the power of big data. \n\nIn th is session\, I will showcase the basic building blocks of storytelling acr oss different mediums - oral storytelling\, journalistic written stories\, graphic comics and movies. And then explore the idea that we can integera te narrative storytelling lessons from these mediums with our data visuali sation to start crafting visual stories with data. \n\nI will summarize ba sic design principles that can help us in our crafting journey\, as we tak e the data through the layers of abstraction - See the Data | Show the Vis ual | Tell the Story | Engage the Audience. The focus would be on sharing 'why' stories work and aim to unpack the six dimensions of creating a data -visual-story.\n \n1. Abstraction (data patterns)\n2. Framing (perspectiv e\, genre)\n3. Representation (visual encoding including color)\n4. Messag ing (verbal\, text annotation)\n5. Flow (arrangement\, transition)\n6. Int eractivity\n\nI will be using exemplars from my work and other real-world data-stories to explore this topic.\n\n### Speaker bio\n\nI am interested in learning and teaching the craft of telling visual stories with data. I use storytelling and data visualization as tools for improving communicati on\, persuasion and leadership. I am a partner at [narrativeviz Consulting ](http://narrativeviz.com) where I conduct workshops and trainings for cor porates\, non-profits\, colleges\, and individuals. I also teach sessions on storytelling with data as invited expert / guest faculty in data visual ization and analytics related courses at IIM Bangalore and IIM Ahmedabad. My background is in strategy consulting in using data-driven stories to dr ive change across organizations and businesses. I have more than 12 years of consulting experience\, first with AT Kearney in India and then with Bo oz & Company in Europe. I did my B.Tech from IIT\, Delhi and PGDM from IIM \, Ahmedabad. You can find more about me at http://amitkaps.com and tweet me at @amitkaps\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/crafting-visual-storie s-with-data-UgTNaWHpRRoyg8Sjtt42xr BEGIN:VALARM ACTION:display DESCRIPTION:Crafting Visual Stories with Data in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Flash talks DTSTART:20140726T091500Z DTEND:20140726T094500Z DTSTAMP:20260421T123212Z UID:session/8HvJkH4EmB1SomYkoLwcPa@hasgeek.com SEQUENCE:0 CREATED:20140705T030352Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140721T051246Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Flash talks in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Flash talks DTSTART:20140726T091500Z DTEND:20140726T100000Z DTSTAMP:20260421T123212Z UID:session/BzpmRMNaq7FVJ9YLReKf2N@hasgeek.com SEQUENCE:0 CREATED:20140721T051232Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140721T051256Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Flash talks in Auditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Real Time User-Scoring for Bidding in Display Retargeting DTSTART:20140726T094500Z DTEND:20140726T100000Z DTSTAMP:20260421T123212Z UID:session/DtsEqkAxm5hqpscQJHonb5@hasgeek.com SEQUENCE:2 CATEGORIES:Crisp talk,Beginner CREATED:20140705T030152Z DESCRIPTION:To participate in an auction\, we must come up with a dollar v alue that we are ready to bid for each user. To do this\, we analyze the s ite-activity of the users in real-time\, along with any meta-data we may h ave. In this session\, we walk through how we have broken this problem int o smaller pieces\, and through optimizations based on CTR\, conversion pro bability and expected revenue\, our campaigns are designed towards achievi ng business goals measured in terms of achieved Return Over Ad Spend (ROAS ). We also handle ad-serving\, and deciding items to recommend in the ad u nits\, in case of bids won.\n\n### Speaker bio\n\nI\, Ambuj Singh\, gradua ted in Computer Science from IIT Kanpur in 2012. Since then\, I have been working in @WalmartLabs\, first in the twitter analysis team\, and current ly in the Display Ads team.\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/real-time-user-scoring -for-bidding-in-display-retargeting-DtsEqkAxm5hqpscQJHonb5 BEGIN:VALARM ACTION:display DESCRIPTION:Real Time User-Scoring for Bidding in Display Retargeting in A uditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Evening tea break DTSTART:20140726T100000Z DTEND:20140726T103000Z DTSTAMP:20260421T123212Z UID:session/JuStfEjHgSoYinjZu5x7YP@hasgeek.com SEQUENCE:0 CREATED:20140527T070132Z DESCRIPTION:\n LAST-MODIFIED:20140703T191341Z ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Evening tea break in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Application of analytics in retail - Anindo Chakraborthy DTSTART:20140726T103000Z DTEND:20140726T111500Z DTSTAMP:20260421T123212Z UID:session/5hG3fZ7VCZ8gHqsTjtxDbm@hasgeek.com SEQUENCE:0 CREATED:20140724T135511Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140724T135523Z LOCATION:Auditorium 2 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Application of analytics in retail - Anindo Chakraborthy in Au ditorium 2 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Experimentation to Productization : developing a Dynamic Bidding s ystem for a location aware Mobile landscape DTSTART:20140726T103000Z DTEND:20140726T111500Z DTSTAMP:20260421T123212Z UID:session/Hmp28v3Z3DZByPCmkhNTCw@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140705T030249Z DESCRIPTION:I will start with the obvious\, ie. how do advertisers reach y ou on mobile\, fetching your comprehensive digital footprint\, all in 6 mi lliseconds\, or less. Then look through sample digital footprint(weblogs)\ , laying the ground to understand the data\, and algorithms to derive stat istical relationships. \n\nTowards this\, I will then talk about identifyi ng quick wins to deliver outcomes\, all through data and in this introduce Hypothesis based Engineering - ie how not to go down a bottomless pit.\n\ nI will then spend majority of time talking about 3 problems we adressed a t AdNear to increase the bottom line for our clients -\n\n1. Algorithm to develop a Dynamic bidding system which prices each opportunity to bid\, ba sed on the quality of that "specific" inventory - Towards this\, I will fo cus on how we built meta-data for otherwise\, not so userful attributes li ke "user-agent"\, data for creatives\, besides the obvious "features"\n\n2 . Characterizing user-mobilty patterns to generate user profiles - ie give n a cross section of user\, how do we map the activities associated with t heir geographical footprint - and generate & probablistic picture of his a ctivity patterns & affinity towards general activities\n\n3. Developing a comprehensive app-ranking system : How we use web to increase the informat ion content of the apps to deliver Business outcomes that matter. The syst em updates the snapshot across multiple dimensions for each of the unique appids in the system every hour\, to deliver a self aligning machine learn ing system at scale\n\nFinally I will close this with the framework we bui lt to measure all this in real time - A/B testing framework\, Simulation & Reporting - which supported the Experimentation phase\, created stickness that pushed the productization of data into our Production systems\, whil e doing so at Scale.\n\n### Speaker bio\n\nEkta is Data Scientist with AdN ear Pte.\, where she is designing Dynamic Bidding systems and A/B testing framework for bidding in location based mobile targeting space\, to increa se the bottom line for clients across Asia-Pacific. She has a background i n Quantitative Economics(MS) from Goethe-University\, Frankfurt and Comput er Science(BS) from Bangalore\, India and enjoys Monetizing and leveraging technology to solve abstract Business problems. While at Grad school she became passionately interested in rationality\, framing problems and how w e human being respond to ambiguous choices\, something she sews in technic al dimensions with a scientific rigour.\n\nPrior to AdNear\, she was with [24]7 Inc.\, Innovation Labs\, where she was responsible for end to end so lutioning\, statistical analysis and deployment of Analytic models for e-c ommerce clients and designing intuitive customer experiences. Before that she has worked in roles across Quality Engineering (VMware Inc.)\, Program Management (SAP Labs) and Experimentation methods\, Auctions & Macroecono mics while pursuing her Masters at Goethe University.\n\nShe presented a t alk at Pycon 2013\, Bangalore\, selected as a speaker for Pycon APAC\, 201 4 Taipei. Also accepted to present the same in Grace Hopper's conference f or Women in Computing\, 2014 at Pheonix(USA)\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/experimentation-to-pro ductization-developing-a-dynamic-bidding-system-for-a-location-aware-mobil e-landscape-Hmp28v3Z3DZByPCmkhNTCw BEGIN:VALARM ACTION:display DESCRIPTION:Experimentation to Productization : developing a Dynamic Biddi ng system for a location aware Mobile landscape in Auditorium 1 in 5 minu tes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:The ART of Data Mining - Practical Learnings from Real-world Data Mining applications DTSTART:20140726T111500Z DTEND:20140726T120000Z DTSTAMP:20260421T123212Z UID:session/C1cLLT5Z24fPr3XhZfAQBV@hasgeek.com SEQUENCE:2 CATEGORIES:Full talk,Intermediate CREATED:20140705T030304Z DESCRIPTION:The role of a data scientist has evolved in the last few years from someone who can "put-together" a "modelling pipeline" to someone who can: (a) "understand" the data beyond basic statistics and simple visuali zations\, (b) extract "deep" and "novel" insights from the data\, (c) engi neer "better features" to fairly distribute complexity between features an d models\, (d) visualize and make sense of complex data types like network s\, unstructured text corpora\, etc.\, and (e) create innovative ways of h arnessing data to make smarter decisions.\n\nIn order to create "magic fro m data"\, a data scientist must go beyond the SCIENCE\, ENGINEERING\, and PROCESS and delve into the ART of data mining. In this talk I will share a number of "mistakes" and "innovations" in this context that helped me bui ld better models in domains as diverse as remote sensing\, text classifica tion\, text clustering\, fraud detection\, information retrieval\, bioinfo rmatics\, retail data mining\, and image understanding\, etc. \n\nThese pr actical insights might help the audience pay attention to the right detail s in the modelling process\, look for model improvements in the right plac es\, be more creative with their data and use its full potential\, and eve n overcome the limitations of their modeling tools.\n\n### Speaker bio\n\n http://www.linkedin.com/in/shaileshk\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20230810T072606Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com URL:https://hasgeek.com/fifthelephant/2014/schedule/the-art-of-data-mining -practical-learnings-from-real-world-data-mining-applications-C1cLLT5Z24fP r3XhZfAQBV BEGIN:VALARM ACTION:display DESCRIPTION:The ART of Data Mining - Practical Learnings from Real-world D ata Mining applications in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:Feedback DTSTART:20140726T120000Z DTEND:20140726T121500Z DTSTAMP:20260421T123212Z UID:session/QkwX63qTco5syPVrADjTbe@hasgeek.com SEQUENCE:0 CREATED:20140705T030502Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20140705T030520Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:Feedback in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT BEGIN:VEVENT SUMMARY:HasGeek-Nexus powered panel on "Real-time analytics – technolog ies of today\, for tomorrow" DTSTART:20140726T121500Z DTEND:20140726T134500Z DTSTAMP:20260421T123212Z UID:session/SZhQn9PD8xADSxkoFRbk2T@hasgeek.com SEQUENCE:0 CREATED:20140630T185331Z DESCRIPTION:\n GEO:12.943181274247097;77.59629179723562 LAST-MODIFIED:20200619T062516Z LOCATION:Auditorium 1 - NIMHANS Convention Centre\nBangalore\nIN ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com BEGIN:VALARM ACTION:display DESCRIPTION:HasGeek-Nexus powered panel on "Real-time analytics – techn ologies of today\, for tomorrow" in Auditorium 1 in 5 minutes TRIGGER:-PT5M END:VALARM END:VEVENT END:VCALENDAR