WORKSHOP
Agentic Analytics in Regulated Industries: Live on Exasol (WS09)
Date: 08 October 2026 | Time: 11:30-02:00 PM
Venue: Workshop Room 3, NIMHANS Convention Centre, Bangalore
FEES:
• Rs.99 for Leaders Pass holders
• Rs.669 for Professionals Pass holders
• Rs.749 for Knowledge Pass holders
• Rs.899 for Community++ or Student Pass holders
• Rs.999 for all others
(Limited seats available)
Show what it takes to put an AI agent on top of enterprise data in industries where every answer may be audited. Using two live builds on public data — a clinical trial landscape assistant for pharma and a near-real-time fraud-scoring pipeline for banking — we demonstrate an agent that discovers schema over MCP, applies the organisation’s definitions through a semantic layer, runs inference inside the database, and leaves a query log that can be replayed months later. The aim is that attendees leave knowing the difference between an agent that has been handed database access and a database built for agents.
Prathamesh Karmalkar
Staff Software Engineer, Exasol AG
- Anyone building or evaluating AI agents over real data — data engineers, backend developers, analytics engineers, architects, and data scientists moving from notebooks to production. No prior experience with Exasol, MCP or the pharma and banking domains is needed; both use cases are explained in plain terms before each demo. It also suits engineering leads and product owners who need to understand what "agentic analytics" actually requires before committing to a platform. Students and early-career developers with basic SQL and Python are welcome.
- Why agentic analytics — what changes when an agent, not a dashboard, answers the question
- What an agent needs from its database: schema discovery, semantic definitions, identity to the row filter, inference inside the boundary, a replayable log
- Exasol in brief — the engine, the open-source MCP server, agent skills, Personal edition
- Demo 1 (pharma): clinical trial landscape assistant over ClinicalTrials.gov — load, semantic view, MCP agent, cited answers, AI functions over eligibility text, audit log
- Demo 2 (banking): near-real-time fraud scoring — Kafka ingestion, in-database scoring UDF, the same agent querying the live stream, row-level policy applied live
- Where this approach doesn't save you
- Take-home kit and Q&A
Benefits/Takeaways of this workshop for the attendees (What will attendees do after attending the workshop which they were not able to do before attending this)
A working mental model of agentic analytics that goes beyond “connect an LLM to a database.” Attendees see the five properties that separate a production-grade setup from a demo, and watch each one enforced live rather than described. They leave with a public GitHub repository containing both builds, the public datasets as snapshots, and a one-command local install of Exasol Personal with the MCP server pre-wired, so everything shown can be reproduced at home. They also get an honest view of the limits — what in-database agents cannot do, and where other tools remain the right choice.
What you’ll need
- At least 8 GB RAM (16 GB is better) and 20 GB free disk space.
- Python 3.11 or newer.
- Internet access to github.com, objects.githubusercontent.com, pypi.org, files.pythonhosted.organd astral.sh. If you’re behind a company firewall or proxy, check these aren’t blocked.
- A SQL client for browsing tables: either VS Code, where setup adds an extension for this, or DbVisualizer, free version: https://www.dbvis.com
- If you’d like to use an AI assistant with the database, have one installed and signed in: Claude Desktop, Claude Code, Cursor, GitHub Copilot in VS Code, Codex, Gemini CLI, OpenCode or Continue.
Windows
- Windows 10 or 11 on an Intel or AMD processor. Windows on ARM (Snapdragon, Copilot+ PCs) won’t work.
- Install Podman Desktop from https://podman-desktop.io or with winget install RedHat.Podman. It may need admin rights and a restart. Then create the Podman machine and leave it rootless, which is the default. If yours is already rootful, run podman machine set –rootful=false.
macOS
- Apple Silicon and Intel Macs both work, with nothing to install first.
- If your company blocks virtualisation on your Mac, it won’t work, so please bring a machine you control.
Linux
- You’ll need Podman (rootless is fine). If it’s missing, setup installs it using sudo, so you’ll need sudo rights.
On a work laptop?
Forward this to your IT team now:
- Access to the five domains above
- Admin rights to install Podman (Windows) or sudo (Linux)
- Virtualisation allowed
- PowerShell execution policy set to RemoteSigned (Windows)
Questions, or stuck on any of this? Write to plg.connect@exasol.com. And if your laptop still won’t cooperate on the day, come anyway; everything is shown on screen and you can follow the whole session.
About Speakers
Prathamesh Karmalkar is a Staff Software Engineer at Exasol AG with over 12 years of experience in AI, Generative AI, NLP, LLMs, machine learning, and deep learning. Recognised among NASSCOM Makers Honor 2026’s Top 30 Engineers of India and as one of the Top 7 AI Scientists at DataHack Summit 2026, he has published 10+ papers in peer-reviewed journals and presented at international conferences. As an AI Product Owner, he focused on building LLM-powered solutions that enable informed decision-making using secure, enterprise data.