BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//HasGeek//NONSGML Funnel//EN
DESCRIPTION:On AI disrupting analytics
X-WR-CALDESC:On AI disrupting analytics
NAME:The Fifth Elephant 2025 Winter Edition
X-WR-CALNAME:The Fifth Elephant 2025 Winter Edition
REFRESH-INTERVAL;VALUE=DURATION:PT12H
SUMMARY:The Fifth Elephant 2025 Winter Edition
TIMEZONE-ID:Asia/Kolkata
X-PUBLISHED-TTL:PT12H
X-WR-TIMEZONE:Asia/Kolkata
BEGIN:VEVENT
SUMMARY:Check-in\; on the spot registrations
DTSTART:20251204T033000Z
DTEND:20251204T035500Z
DTSTAMP:20260419T170433Z
UID:session/8WDNYf6Ggq9axXaVdaAk6h@hasgeek.com
SEQUENCE:1
CREATED:20251118T061603Z
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251118T061607Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Check-in\; on the spot registrations in Auditorium in 5 minute
 s
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Introduction to the Winter Edition - AI disrupting analytics and d
 ata workflows
DTSTART:20251204T035500Z
DTEND:20251204T041000Z
DTSTAMP:20260419T170433Z
UID:session/EMocik83PNgtVv1bFExHu1@hasgeek.com
SEQUENCE:2
CREATED:20251118T061639Z
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251118T061651Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Introduction to the Winter Edition - AI disrupting analytics a
 nd data workflows in Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Inside DataGaaru - Razorpay's self-serve AI analyst
DTSTART:20251204T041000Z
DTEND:20251204T044500Z
DTSTAMP:20260419T170433Z
UID:session/9FyqEn3q8Gt3V1HaoZoyt@hasgeek.com
SEQUENCE:7
CATEGORIES:30 mins talk
CREATED:20251118T062436Z
DESCRIPTION:At Razorpay\, the growing demand for data & insights was incre
 asing manual dependence\, slowing down critical business decisions. This l
 ed us to ask a question: "Are chatbots the future of analytics?" Our answe
 r is DataGaaru\, a home-grown AI Analyst Assistant that is fundamentally r
 eshaping how we interact with data. Today\, it handles over 1\,500 queries
  monthly and has already reclaimed 20% of our analytics team's bandwidth.\
 n\nThis session is a behind-the-scenes case study of our journey building 
 DataGaaru from the ground up. We’ll unpack the architecture of its two c
 ore features: a Reasoning Bot (our text-to-SQL engine) and a Health Metric
 s Document Generator that automates metric deep-dives.A key focus will be 
 our structured prompting strategy\, which injects crucial\, domain-specifi
 c context and business rules into the LLM to achieve high query accuracy.\
 n\nKey Takeways\n\nThis session provides a practical playbook for anyone b
 uilding and scaling AI-driven data tools.\n- Our Blueprint for an AI Analy
 st: A simple walkthrough of our architecture\, how we give the LLM the rig
 ht context\, and our plans to use multiple models to optimize for cost and
  performance.\n- From Idea to Essential Tool: How we rolled it out in phas
 es and focused on basics like security and data logging. We'll show the ex
 act OKRs we used to track our progress\, like cutting down query time from
  4 minutes to under 20 seconds.\n- How We Built Trust: An AI tool is usele
 ss if people don't trust it. We’ll share the simple features we added to
  build confidence\, like a "test questions" library\, a SQL explainer\, an
 d a "Confidence Score" for every query.\n\nProposed Session Format: Presen
 tation + Live Product Demo\n\nIdeal Audience\n- Data Leaders and Managers 
 who want a proven strategy to solve the data and insights bottleneck.\n- D
 ata Product Managers who want to learn how to build and scale a tool users
  trust.\n- Data Analysts who want to see how AI can help them focus on mor
 e important work.\n  \nSpeaker's bio\nAs Razorpay's first analytics hire\,
  Prerit built the Analytics Center of Excellence (COE) from the ground up\
 , scaling analytics and AI across multiple product lines.\n\nLink: https:/
 /drive.google.com/file/d/1ynT-zmfDVQMKACSaNLiD6Pm-XCT2Ca-B/view?usp=drives
 dk\n\n
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251124T012447Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/from-bottleneck
 -to-self-serve-how-razorpay-built-and-scaled-its-ai-analyst-datagaaru-9Fyq
 En3q8Gt3V1HaoZoyt
BEGIN:VALARM
ACTION:display
DESCRIPTION:Inside DataGaaru - Razorpay's self-serve AI analyst in Auditor
 ium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Break
DTSTART:20251204T044500Z
DTEND:20251204T045500Z
DTSTAMP:20260419T170433Z
UID:session/CA4HU3FhJ8PHa59Y71j4oU@hasgeek.com
SEQUENCE:4
CREATED:20251123T062629Z
LAST-MODIFIED:20251124T012449Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:OpenMetadata - making AI smarter with enterprise data context
DTSTART:20251204T045500Z
DTEND:20251204T053000Z
DTSTAMP:20260419T170433Z
UID:session/5tjNRz8UiLKe2fiHFqEVk7@hasgeek.com
SEQUENCE:13
CATEGORIES:30 mins talk
CREATED:20251119T055409Z
DESCRIPTION:AI can generate code\, queries\, and dashboards but it still d
 oesn’t understand your data. Without context\, lineage\, ownership\, and
  business meaning\, AI systems remain disconnected from the reality of you
 r enterprise data.\n\nOpenMetadata bridges this gap by bringing semantic u
 nderstanding to your data \, transforming fragmented assets across databas
 es\, pipelines\, dashboards\, and models into a connected knowledge graph.
  This unified graph gives AI the context it needs to reason about data\, a
 utomate workflows\, and deliver trustworthy insights.\n\nIn this session\,
  we’ll explore how organizations are using OpenMetadata as the semantic 
 layer for AI\, enabling intelligent automation across data governance\, qu
 ality\, and discovery. We’ll also demonstrate how AI agents built on the
  Model Context Protocol (MCP) can act safely and autonomously on metadata 
  optimizing pipelines\, improving quality\, and keeping data systems in sy
 nc.\n\n\n## Key Takeaways\n- Learn how OpenMetadata provides the semantic 
 foundation that allows AI to understand and act on your data.\n- See how k
 nowledge graphs and embeddings make data workflows context-aware and self-
 optimizing.\n- Discover how AI agents powered by OpenMetadata and MCP are 
 driving a new era of automated governance and analytics.\n\n## Ideal Audie
 nce\n- Data and AI leaders\, platform engineers\, and analytics practition
 ers building the next generation of AI-ready data platforms.\n\n## Spekaer
 's Bio\nMayur Singal is Open Source contributor to the OpenMetadata projec
 t\, where he leads the development of connector integrations and lineage.\
 n\nElevator Pitch Deck -https://docs.google.com/presentation/d/1WO0vmYXGJ8
 qOchBN7eiXYnP0LOpmpv1PA-G6VhlPvTk/edit?slide=id.g3a04b90a59a_0_86#slide=id
 .g3a04b90a59a_0_86 \n\n\nMCP Demo -\nhttps://www.youtube.com/watch?v=5FqK3
 Cr9fXI
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251124T012450Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/openmetadata-br
 inging-semantic-understanding-to-your-data-to-power-ai-5tjNRz8UiLKe2fiHFqE
 Vk7
BEGIN:VALARM
ACTION:display
DESCRIPTION:OpenMetadata - making AI smarter with enterprise data context 
 in Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Morning break
DTSTART:20251204T053000Z
DTEND:20251204T060000Z
DTSTAMP:20260419T170433Z
UID:session/Ec2AmhgvHC3eqv1di6GETc@hasgeek.com
SEQUENCE:8
CREATED:20251123T062643Z
LAST-MODIFIED:20251124T012457Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Morning break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:The causality trap: why your LLM sounds smart but thinks shallow
DTSTART:20251204T060000Z
DTEND:20251204T064000Z
DTSTAMP:20260419T170433Z
UID:session/Pz2dsXJqAYvjLzTjJJatyu@hasgeek.com
SEQUENCE:6
CATEGORIES:30 mins talk
CREATED:20251123T062735Z
DESCRIPTION:\nThe talk will start with an introduction / motivation about 
 causal inference in general. Humans have excellent causal models of the wo
 rld\, so associational analytics suffice in human businesses because the c
 ausal model of the business is implicitly known to everyone involved\, tha
 ts why - despite almost no one bothering with causal inference - most BI d
 epartments are doing just fine. On the other hand\, there are situations w
 here knowing the precise flow of causality determines what intervention is
  necessary and what form it takes. With this background\, the talk will in
 troduce Pearl's ladder of causality\, which will form the backbone of the 
 rest of the talk. All of this should take ~ 10 minutes.\n\nLLMs are wonder
 ful at associational reasoning\, but they are working off the average wisd
 om of the internet and will pretend to be causal (because humans habve fil
 led the internet with casually causal language) but lack both the implicit
  model of your business that your mind possesses\, as well as the tools to
  correct themselves. There are benchmarks that show that LLMs get progress
 ively worse as one climbs the ladder of causation. Up to this point\, I sh
 ould be at the 15-17 minute mark. Then\, the talk will quickly walk throug
 h a toy example of causal analysis in R using the wonderful ecosystem ther
 e up to the second rung of the ladder. ~10 more minutes. Leaving - ideally
  - 3-4 minutes for questions.\n\n**Main takeaways :** \n**(a)** The langua
 ge of causality.. what is the ladder\, what are interventions\, counterfac
 tuals\, DAGs. \n**(b)** Causal inference is wonderful\, worth learning\, a
 nd LLMs can't do it out of the box so your agent will need explicit causal
  tools to reason about interventions and counterfactuals. \n\n**Target Aud
 ience :** this is probably data scientists and data science managers who a
 re dealing with non trivial and novel problems and want to see if new tool
 s might be brought to bear on them.\n\n**Bio :** I am a lapsed physicist w
 ith several years of experience in various Data Science and BI contexts\, 
 mostly in Germany. Now I am the founder of Romulan AI - building the causa
 l layer for your LLM first business. 
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251123T065156Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/your-causal-par
 rot-might-be-lying-to-you-Pz2dsXJqAYvjLzTjJJatyu
BEGIN:VALARM
ACTION:display
DESCRIPTION:The causality trap: why your LLM sounds smart but thinks shall
 ow in Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Birds of Feather (BOF) session: AI disrupting data viz
DTSTART:20251204T064000Z
DTEND:20251204T074000Z
DTSTAMP:20260419T170433Z
UID:session/Y3X76h2oxcCmstgrve4nHL@hasgeek.com
SEQUENCE:6
CATEGORIES:Birds of Feather (BOF) session
CREATED:20251126T062009Z
DESCRIPTION:This is a BoF (i.e. interactive\, round table) session bringin
 g together data visualization and AI practitioners\, discussing about how 
 AI is in the process of upending data visualization. \n\nNow it is easy fo
 r non-technical people to easily produce data visualizations\, and communi
 cate effectively - even without using basic tools such as MS Excel. With a
  few prompts\, one can pull and manipulate \ndata and make beautiful chart
 s with it. \n\n**Key Takeaways**\n\nIn this BoF\, we intend to discuss=: \
 n1. Sets of tools / prompts that can help make superior visualizations\, a
 nd communicate effectively.\n2. If\, and how\, we can do so much more in t
 erms of data visualization thanks to the tools that LLMs offer. \n3. Pitfa
 lls\, if any\, of easy access to data viz tools - does this mean we get mo
 re bad visualizations\, and if so\, what we can do to make sure we interpr
 et them properly? \n\n**Intended Audience** \nPractitioners of data visual
 ization\, of all levels (beginner / intermediate / advanced) \n\n**Speaker
  Bio** \n\n**Karthik Shashidhar** is cofounder and CEO of [Babbage Insight
 ](https://www.babbageinsight.com). He is an award winning data science lea
 der\, who was previously the head of analytics and data science at Delhive
 ry. He has been a [columnist for Mint](https://www.livemint.com/authors/ka
 rthik-shashidhar)\, and has written [*Between The Buyer And The Seller*](h
 ttps://www.amazon.in/Between-Buyer-Seller-Karthik-Shashidhar-ebook/dp/B074
 WC3FFX)\, a book on market design. He has a lifetime corpus of 3000+ blog 
 posts\, and he now writes at [Pertinent Observations](https://noenthuda.su
 bstack.com). \n
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251126T062100Z
LOCATION:Birds of Feather - Room 1 - Samarthanam Auditorium\nBengaluru\,\n
 IN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/ai-disrupting-d
 ata-viz-Y3X76h2oxcCmstgrve4nHL
BEGIN:VALARM
ACTION:display
DESCRIPTION:Birds of Feather (BOF) session: AI disrupting data viz in Bird
 s of Feather - Room 1 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Break
DTSTART:20251204T064000Z
DTEND:20251204T065000Z
DTSTAMP:20260419T170433Z
UID:session/D4L76CEBsEB3o3t93rsnmH@hasgeek.com
SEQUENCE:1
CREATED:20251123T062834Z
LAST-MODIFIED:20251123T062836Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Synthetic data at scale: enterprise-ready workflows for privacy & 
 compliance
DTSTART:20251204T065000Z
DTEND:20251204T072500Z
DTSTAMP:20260419T170433Z
UID:session/DZdSaz5Q1HNQZ1kXg9Hqf3@hasgeek.com
SEQUENCE:7
CATEGORIES:15 mins talk
CREATED:20251123T062912Z
DESCRIPTION:In AI driven workflows\, a critical bottleneck is the scarcity
  of realistic training datasets—and strict privacy and security rules fo
 rbid using actual customer records as well.  To address these challenges\,
  we have developed an agentic synthetic data generation pipeline that prod
 uces domain-rich\, realistic\, and coherent datasets\, for training PII (P
 ersonally Identifiable Information) and sensitive-reference detection mode
 ls\, preserving customer’s privacy. This LLM-driven workflow autonomousl
 y curates synthetic samples across varied industries—such as finance\, h
 ealthcare\, and legal—while incorporating guardrails to ensure that gene
 rated content remains non-toxic\, unbiased\, and contextually safe.  \n\nI
 n this session\, we will present NetApp’s end-to-end framework for detec
 ting sensitive data and sensitive references\, powered by synthetic data. 
 Our approach demonstrates how synthetic datasets can effectively bridge th
 e data availability gap while maintaining strong alignment with real-world
  linguistic patterns. Through extensive experimentation\, we observed prog
 ressive improvements in detection accuracy as synthetic data volume and di
 versity increased. The session will delve into the architecture of the age
 ntic pipeline\, data quality validation strategies\, and domain adaptation
  techniques. Attendees will gain insights into how synthetic data can enab
 le responsible AI development\, reinforce data governance\, and ensure com
 pliance without exposing or relying on real customer information.\n\n## Ta
 keaways:\n1. Real-world data scarcity no longer bottlenecks model training
  or fine-tuning—high-quality synthetic corpora can fill the gap. Diverse
 \, coherent synthetic datasets are key to achieving robust\, generalizable
  performance across domains.\n2. By leveraging agentic synthetic-data gene
 ration\, we create datasets that so closely mimic real-world documents the
 y’re indistinguishable from genuine records—and we’ve observed consi
 stent performance improvements with each increment of quality synthetic sa
 mples\, motivating continued investment in this approach\n\n## Target audi
 ences\nThis session will be particularly beneficial for machine learning e
 ngineers\, data scientists\, and AI researchers working on privacy-sensiti
 ve applications or responsible AI initiatives. It will also provide valuab
 le insights for leaders/architects working in sensitive or high security d
 omains where data governance and compliance play important role. Attendees
  from organizations dealing with regulated data—such as finance\, health
 care\, and government sectors—will gain an understanding of how syntheti
 c data can be strategically leveraged to enhance model performance while m
 aintaining strict privacy guarantees. \n\n\n### Authors\n**Presenter**: \n
 [Darshan Adiga](https://www.linkedin.com/in/d-adiga/)\, \nSenior Data Scie
 ntist at NetApp\n\n\n**Co-author**:\n[Lakshya Daulani](https://www.linkedi
 n.com/in/lakshyadaulani/)\,\nData Scientist at NetApp\n
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251124T012300Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/enterprise-read
 y-compliant-synthetic-data-generation-for-data-governance-DZdSaz5Q1HNQZ1kX
 g9Hqf3
BEGIN:VALARM
ACTION:display
DESCRIPTION:Synthetic data at scale: enterprise-ready workflows for privac
 y & compliance in Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Lunch break
DTSTART:20251204T072500Z
DTEND:20251204T082000Z
DTSTAMP:20260419T170433Z
UID:session/5W7pgcyP4f7cHJs1vSeEac@hasgeek.com
SEQUENCE:3
CREATED:20251118T061851Z
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251123T062923Z
LOCATION:Birds of Feather - Room 1 - Samarthanam Auditorium\nBengaluru\,\n
 IN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Lunch break in Birds of Feather - Room 1 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Birds of Feather (BOF) session: Stories and guides - data governan
 ce & security guardrails that actually work
DTSTART:20251204T081000Z
DTEND:20251204T091000Z
DTSTAMP:20260419T170433Z
UID:session/4RsCx2LVvRdpNWfJyf22xj@hasgeek.com
SEQUENCE:3
CATEGORIES:Birds of Feather (BOF) session
CREATED:20251126T061854Z
DESCRIPTION:AI has augmented and changed how analytics teams ingest data\,
  generate insights\, and build decision systems. But as AI-driven analytic
 s becomes more agentic i.e. writing language to SQL\, accessing datasets\,
  generating dashboards\, and triggering workflows\, achieve simple to comp
 lex outcomes\, the risk surface increases and needs some controls. \nPoor 
 guardrails can lead to:\na) Unauthorized data exposure\nb) Cost explosions
  from unbounded queries\nc) Silent model drift impacting business decision
 s (resulting in bias?)\nd) Incorrect insights generated with high confiden
 ce\ne) Compliance violations during AI-initiated data operations\n\nThis B
 oF session brings practitioners together to answer one question:\nHow do w
 e make AI-driven analytics safe. Without breaking governance\, trust\, or 
 compliance.\nThis is not a talk or a lecture. This is real world sharing. 
 Patterns\, failures\, guardrails which are being used in production analyt
 ics environments in enterprises.\n\n\nKey Takeaways\n-\n1. Navigator/bluep
 rint for designing security & governance guardrails for AI-driven analytic
 s (data access to insight validation).\n2. Community understanding of risk
 s unique to LLM based analytics (semantic model drift\, hallucinated joins
 \, PII leakage\, rogue queries).\n3. Tribal knowledge (since the practices
  are evolving) for safe adoption of AI copilots\, agents\, and NLP-based a
 nalytics interfaces.\n4. Strategies to ensure analytics teams benefit from
  AI speed without sacrificing trust\, compliance\, or cost control.\n\nWho
  will benefit\n-\n\nData Platform / Analytics Engineering Teams (hands-on 
 with warehouses\, catalogs\, and BI systems)\nEngineers or Teams building 
 co-pilot systems (LLM-based workflows\, agents\, semantic layers\, and mul
 ti-step/hybrid systems)\nCompliance and Risk Teams (working with regulator
 y\, process for handling regulated or sensitive data)\n \n\nBOF organizer\
 n-\nRavi Balgi is a Architect at Datanimbus with 15 years of experience in
  designing and building large data driven applications within enterprises 
 across multiple domains. Last 8 years have been spent on building large re
 altime data applications\, migrating from legacy systems
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251126T061949Z
LOCATION:Birds of Feather - Room 1 - Samarthanam Auditorium\nBengaluru\,\n
 IN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/bof-stories-and
 -guides-data-governance-security-guardrails-that-actually-work-4RsCx2LVvRd
 pNWfJyf22xj
BEGIN:VALARM
ACTION:display
DESCRIPTION:Birds of Feather (BOF) session: Stories and guides - data gove
 rnance & security guardrails that actually work in Birds of Feather - Room
  1 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Stop querying tables\; give AI governed metrics instead
DTSTART:20251204T082000Z
DTEND:20251204T090000Z
DTSTAMP:20260419T170433Z
UID:session/2Eyav8d2Fj3VmTGoB17t8S@hasgeek.com
SEQUENCE:6
CATEGORIES:30 mins talk
CREATED:20251124T012411Z
DESCRIPTION:As AI becomes central to analytics workflows\, many teams try 
 to let LLMs query databases directly—only to find that raw tables lack t
 he semantics AI needs. The result: hallucinated KPIs\, ambiguous joins\, i
 nconsistent definitions\, invalid SQL\, and costly full-table scans.\n\nTh
 is talk presents a better approach: a SQL-based metrics layer that acts as
  the semantic boundary between AI and data. We’ll explore why metrics—
 not tables—are the smallest meaningful unit of analytical reasoning\, an
 d why SQL is the ideal language for defining them. We’ll look at the whi
 ch OLAP engines can be used to enable sub-second metric iteration and how 
 the metrics layer becomes the foundation for conversational analytics.\n\n
 \nAttendees will learn best practices for metric governance\, security and
  access control\, AI-assisted modeling\, and preventing ungoverned KPI cre
 ation. We’ll introduce **Rill’s metric-based semantic architecture**\,
  security policies\, and MCP tools that allow AI agents to safely discover
 \, query\, and explain governed metrics—without ever exposing raw tables
 .\n\nA live demo will close the session\, showing how AI can reliably answ
 er complex business questions using governed metrics with speed\, determin
 ism\, and explainability.\n\n### **You’ll walk away with:**\n- A bluepri
 nt for a SQL-based metrics layer that both AI and humans can interact with
 \n- Best practices for governance\, versioning\, and KPI approval workflow
 s\n- A deeper understanding requirements of conversational analytics\n- St
 rategies for securing data access when AI agents are in the loop\n- Techni
 ques for using AI responsibly in metric design and modeling\n- A working e
 xample of how Rill’s SQYAML + MCP tooling creates safe\, reliable AI wor
 kflows\n\n### Bio: \nNishant Bangarwa is the Co-Founder and Head of Engine
 ering at Rill Data\, where he leads the development of Rill’s open-sourc
 e\, high-performance\, AI-native BI application—an alternative to legacy
  business intelligence tools\, powered by modern analytical databases and 
 a BI-as-Code workflow that delivers sub-second insights at scale.\n\nPrior
  to Rill\, Nishant worked on large-scale analytics systems at Cloudera and
  Metamarkets\, and has contributed to several major open-source projects\,
  including Apache Druid\, Apache Calcite\, and Apache Hive. His work has f
 ocused on advancing the state of the art in data infrastructure\, real-tim
 e analytics\, and metric-driven decision systems.\n\n\nHere is a link to e
 levator pitch - \nhttps://docs.google.com/presentation/d/1Krn2XGpgdrZ6Uo9B
 25fnfG5Tyi9hsISZd7QDE836OI0/edit?usp=sharing
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002249Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/stop-prompting-
 your-database-give-ai-a-metrics-layer-instead-2Eyav8d2Fj3VmTGoB17t8S
BEGIN:VALARM
ACTION:display
DESCRIPTION:Stop querying tables\; give AI governed metrics instead in Aud
 itorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Break
DTSTART:20251204T090000Z
DTEND:20251204T091000Z
DTSTAMP:20260419T170433Z
UID:session/AzxT44PGGpHNhbooNSZLDq@hasgeek.com
SEQUENCE:2
CREATED:20251123T063724Z
LAST-MODIFIED:20251129T002250Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Flash talks
DTSTART:20251204T091000Z
DTEND:20251204T093000Z
DTSTAMP:20260419T170433Z
UID:session/Qn3eKhF45rW7u7D2NZxPqG@hasgeek.com
SEQUENCE:4
CREATED:20251118T061944Z
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002251Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Flash talks in Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Break
DTSTART:20251204T093000Z
DTEND:20251204T094000Z
DTSTAMP:20260419T170433Z
UID:session/NbZNRwNhvmXewVphN7FSFH@hasgeek.com
SEQUENCE:3
CREATED:20251124T012354Z
LAST-MODIFIED:20251129T002253Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Decoding 1951 census: LLMs extract messy tables
DTSTART:20251204T094000Z
DTEND:20251204T102500Z
DTSTAMP:20260419T170433Z
UID:session/FysnGvxc3eoFvsn5Wa3JgB@hasgeek.com
SEQUENCE:13
CATEGORIES:30 mins talk
CREATED:20251123T062015Z
DESCRIPTION:Traditional OCR systems are excellent at recognizing text but 
 fall short in understanding the structure of complex historical documents
 —particularly tables with varying column arrangements\, inconsistent lay
 outs\, and annotations. This structural ambiguity often makes it infeasibl
 e to extract reliable\, machine-readable data from large collections of su
 ch documents.\n\nWe encountered these challenges while extracting village-
 level demographic and public goods data from India’s **1951 Population C
 ensus (PC51)** district handbooks. These handbooks contain invaluable gran
 ular data that could unlock decades-spanning research across economics\, d
 emography\, and development. However\, their ad-hoc formatting across stat
 es and complex visual structures made OCR-based approaches unsuitable for 
 systematic data extraction.\n\nWe developed an LLM-based pipeline purpose-
 built for large-scale extraction from such format-variant documents. Our a
 pproach leverages LLMs’ ability to interpret context and infer structura
 l meaning—making it possible\, for the first time\, to extract harmonize
 d microdata from these messy tabular layouts. Crucially\, we focused on _r
 eliability_: combining LLMs with traditional\, rule-based techniques such 
 as manually defined table types\, schema templates\, and a robust evaluati
 on framework to ensure data accuracy and consistency.\n\nThis talk will pr
 esent our technical pipeline architecture\, evaluation methodology\, and k
 ey lessons learned. We’ll share practical principles for designing LLM-b
 ased extraction systems that generalize\, and balance automation with huma
 n-informed rules.\n\n**Key Takeaways:**\n\n*   How we built a reliable LLM
 -powered extraction pipeline for format-variant historical documents\, suc
 h as district handbooks from the 1951 Population Census.\n    \n*   Combin
 ing contextual reasoning with rule-based templates and structured evaluati
 ons.\n    \n*   When to replace or augment traditional methods with LLM-dr
 iven extraction.\n    \n*   What new data sources can be unlocked as these
  models evolve.\n    \n\n**Who Should Attend:**\n\n*   Data practitioners 
 and researchers working with complex\, historical\, or unstructured docume
 nt collections.\n    \n*   Engineers exploring scalable\, reliable LLM wor
 kflows.\n    \n*   Anyone interested in unlocking data trapped in legacy f
 ormats.\n\n**Bio**\nAryan Srivastava is a Data Scientist at [Development D
 ata Lab](https://www.devdatalab.org/). 
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002307Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/extracting-data
 -from-historical-documents-harnessing-llms-to-parse-format-variant-tables-
 at-scale-FysnGvxc3eoFvsn5Wa3JgB
BEGIN:VALARM
ACTION:display
DESCRIPTION:Decoding 1951 census: LLMs extract messy tables in Auditorium 
 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Shaping India’s AI Regulation: A Dialogue on MeitY’s Draft IT 
 Rules on Synthetic Content
DTSTART:20251204T094000Z
DTEND:20251204T104000Z
DTSTAMP:20260419T170433Z
UID:session/QuTbe6mH8L7tMNCniCmm1Z@hasgeek.com
SEQUENCE:4
CREATED:20251203T032418Z
DESCRIPTION:The Indian government has publicly articulated an innovation-f
 irst approach to AI governance\, favouring light-touch\, voluntary framewo
 rks while leveraging amendments to existing laws to address emerging AI ri
 sks. Synthetically generated content and deepfakes\, in particular\, have 
 emerged as the most visible and politically sensitive manifestations of AI
 -related harms. From manipulated videos of public figures in compromising 
 situations to their use in electoral contexts\, deepfakes have drawn susta
 ined regulatory and public scrutiny. In a first instance of regulating AI 
 through existing statutory frameworks in the IT sector\, the Ministry of E
 lectronics and Information Technology (MeitY)\, on 22 October 2025\, publi
 shed draft amendments to the Information Technology (Intermediary Guidelin
 es and Digital Media Ethics Code) Rules\, 2021\, in relation to synthetica
 lly generated information (Draft Amendments). The amendments were open for
  public consultation until 13 November 2025. However\, the Ministry has no
 t yet finalised the framework\, leaving important space for further engage
 ment with the government and for grounded stakeholder inputs to shape the 
 final approach.
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251203T032608Z
LOCATION:Birds of Feather - Room 2 - Samarthanam Auditorium\nBengaluru\,\n
 IN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Shaping India’s AI Regulation: A Dialogue on MeitY’s Draft
  IT Rules on Synthetic Content in Birds of Feather - Room 2 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Exploring the semantic layer underlying your data
DTSTART:20251204T094000Z
DTEND:20251204T104000Z
DTSTAMP:20260419T170433Z
UID:session/BEYphMxQavktfVCGeWqB7a@hasgeek.com
SEQUENCE:8
CREATED:20251124T012720Z
DESCRIPTION:In a world where every team is trying to make sense of explodi
 ng data and fast-moving AI tools\, one big question keeps coming up: how d
 o we actually give our data meaning? Join us for an informal unconference 
 session — “Exploring the Semantic Layer Underlying Your Data” — wh
 ere the group\, not a panel\, drives the conversation.\n\nWe want to look 
 at approaches for how people really contextualize enterprise knowledge. Wh
 at’s working in the wild — knowledge graphs\, shared metrics layers\, 
 domain-driven modeling? What hasn’t lived up to the hype? Expect honest 
 stories\, not polished pitches.\n\nFrom there\, we’ll open the floor to 
 the everyday challenges participants face: messy data landscapes\, conflic
 ting definitions\, and the tension between governance and speed. If you’
 ve struggled to get teams aligned on “the truth\,” you’re in good co
 mpany.\n\nWe’ll also explore how to validate and observe AI-generated an
 swers\, especially as more analytical work flows through natural-language 
 tools. How do you know the answers are right? How do you show your work?\n
 \nFinally\, we’ll talk about building trust with users — transparency\
 , provenance\, and the human factors that matter more than we often admit.
 \n\nCome ready to share\, challenge\, and co-create ideas.
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002255Z
LOCATION:Birds of Feather - Room 1 - Samarthanam Auditorium\nBengaluru\,\n
 IN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Exploring the semantic layer underlying your data in Birds of 
 Feather - Room 1 in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Break
DTSTART:20251204T102500Z
DTEND:20251204T103500Z
DTSTAMP:20260419T170433Z
UID:session/TcczitpCceUoki1cZ9pWB9@hasgeek.com
SEQUENCE:5
CREATED:20251123T063822Z
LAST-MODIFIED:20251129T002311Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Break in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:How Swiggy builds reliable agentic data co-pilots at scale
DTSTART:20251204T103500Z
DTEND:20251204T111500Z
DTSTAMP:20260419T170433Z
UID:session/CkTFrmBrqKUFuhL4XRS5Me@hasgeek.com
SEQUENCE:13
CATEGORIES:30 mins talk
CREATED:20251118T062314Z
DESCRIPTION:Building Data Co‑Pilots is a session on how to use agentic L
 LMs to reliably answer complex business questions by aggregating data acro
 ss many sources—warehouse/lakehouse tables\, metrics/semantic layers\, f
 eature stores\, logs\, and knowledge bases. We will walk through the syste
 m design that turns a plain‑English question (e.g.\, “Why did cancella
 tions spike in City X yesterday?”) into an executable plan: decomposing 
 the task\, selecting tools\, generating and validating SQL\, auto‑joinin
 g via a semantic catalog\, applying statistical checks\, and returning con
 cise narratives with charts and lineage. We will also cover governance‑b
 y‑design: policy enforcement\, PII handling\, cost control\, observabili
 ty\, and human‑in‑the‑loop review.\n\nWe will demo working implement
 ations of Swiggy data co‑pilots that support multi‑source exploration 
 and root‑cause workflows. You’ll see the end‑to‑end flow—from pr
 ompt to plan/DAG\, semantically consistent queries across heterogeneous sc
 hemas\, confidence scoring and guardrails\, to final decision‑ready answ
 ers—plus the playbook we used to move from prototype to production (eval
 uation harnesses\, MCP (Model Context Protocol)‑style adapters for data 
 tools\, caching\, and failure fallbacks).\n\nTakeaways:\n1. A reference ar
 chitecture and design patterns to build reliable agentic analytics co‑pi
 lots over messy\, multi‑source data—plus a production‑readiness chec
 klist.\n\n2. Demo‑backed lessons on what actually works: semantic/metric
 s layer abstraction\, MCP‑style tool adapters\, guardrails and evaluatio
 n\, observability\, and cost governance.\n\nWho Will Benefit:\n1. Data eng
 ineers\, analytics engineers\, BI developers\, and data platform owners\n2
 . Data scientists and ML engineers exploring RAG/agentic patterns for anal
 ytics\n3. Product managers and analytics leaders driving self‑serve insi
 ghts and operational decisioning\n\nWho are we:\nMeghana Negi\, Senior Man
 ager\, Data Science\, Swiggy\nGoda Ramkumar\, Vice President | Head of Dat
 a Science at Swiggy\nAmaresh Marripudi\, Principal Product Manager\, Swigg
 y
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002314Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
URL:https://hasgeek.com/fifthelephant/2025-winter/schedule/building-data-c
 o-pilots-that-work-CkTFrmBrqKUFuhL4XRS5Me
BEGIN:VALARM
ACTION:display
DESCRIPTION:How Swiggy builds reliable agentic data co-pilots at scale in 
 Auditorium in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Transition
DTSTART:20251204T111500Z
DTEND:20251204T112000Z
DTSTAMP:20260419T170433Z
UID:session/QJBXi6ipWapRKHLJVaprzB@hasgeek.com
SEQUENCE:1
CREATED:20251124T012751Z
LAST-MODIFIED:20251129T002317Z
LOCATION:Samarthanam Auditorium\, Bengaluru
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Transition in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
BEGIN:VEVENT
SUMMARY:Feedback\; concluding notes\; bi-monthly meet-ups
DTSTART:20251204T112000Z
DTEND:20251204T114500Z
DTSTAMP:20260419T170433Z
UID:session/E6voMTyZQUtb9PMYt8HJE6@hasgeek.com
SEQUENCE:11
CREATED:20251118T062014Z
GEO:12.9149604;77.6389449
LAST-MODIFIED:20251129T002319Z
LOCATION:Auditorium - Samarthanam Auditorium\nBengaluru\,\nIN
ORGANIZER;CN="The Fifth Elephant":MAILTO:no-reply@hasgeek.com
BEGIN:VALARM
ACTION:display
DESCRIPTION:Feedback\; concluding notes\; bi-monthly meet-ups in Auditoriu
 m in 5 minutes
TRIGGER:-PT5M
END:VALARM
END:VEVENT
END:VCALENDAR