Free Live 45-min Build

Build a Text-to-SQL AI Agent on Databricks — Live, with Cost Guardrails

Deploy db-agent on a real Databricks workspace in 45 minutes — with the data/LLM boundary explicit and the complete cost-guardrail layer installed live. Statement allowlists, query cost checks, retry ceilings, context budgets. Every token-spending decision narrated.

July 8 · 12:00 PM EDT

Reserve your free seat

Microsoft Teams link sent instantly. Replay available if you can't attend live.

Free · No credit card · Teams link sent instantly

July 8, 2026
12:00 – 12:45 PM EDT
Limited seats
What we'll build, live
Why bundled LLM tools (Databricks Genie and its equivalents) cost you governance, billing visibility, and token control — and what to do instead
The architectural principle: separate your data infrastructure from your LLM infrastructure, and own the boundary
Live deployment of db-agent on a real Databricks workspace — Unity Catalog scoping, role-aware access, the full setup end-to-end
The complete cost-guardrail layer installed live: statement allowlists (no DROP / TRUNCATE / UPDATE without explicit approval), query cost checks before execution, retry ceilings, and per-call context budgets
How to make LLM spend its own visible cost center instead of a hidden line inside compute billing
Model routing patterns — when to send a question to a cheap small model vs. a premium large model, and how to enforce it
Live demo: real business questions answered against a sample lakehouse with the full audit trail visible
Live Q&A — bring your data platform questions
C

Chandan Kumar

Founder, beCloudReady · Creator of open-source db-agent (AAAI-25 workshop) · 4,000+ A100/H100 GPU operations background

Why this session, why now

Bundled LLM-over-data tools — Databricks Genie and its equivalents — are convenient and opaque by design. You don't choose the model. You can't route simple questions to a cheaper one. You can't set token ceilings or retry limits. And the AI consumption arrives entangled with platform compute, so when spend climbs, you can't isolate which decision caused it.

The teams that stay in control are the ones that kept the boundary between their data infrastructure and their LLM infrastructure. This session shows you exactly what that looks like in production — built live, in 45 minutes.

Deeper context: The Token Playbook — why dev teams burn AI budgets in a week, and the two disciplines that fix it.

Who this session is for

Data engineers & analytics engineers

You run a Databricks or Snowflake lakehouse and business users want to ask it questions in plain English — but you've seen Genie's cost curve and want a version you can govern, route, and budget.

Platform & security leads

You need an LLM-over-data deployment that respects Unity Catalog, keeps the audit trail explicit at the boundary, and doesn't let an agent run arbitrary SQL against production.

Engineering managers & data leaders

Your AI-on-data line item is opaque and growing. You want LLM spend isolated from platform compute, with token ceilings enforced in architecture — not in a memo.

Engineers building on db-agent

You're evaluating or already using the open-source db-agent and want to see the production deployment patterns and cost guardrails the maintainer uses in real engagements.

AAAI-25 Workshop ProjectOpen Source on GitHub200+ Corporate Training Programs DeliveredDatabricks Partner

For data teams

Want the full version on your own Databricks workspace?

The same build, done by your engineers on your Databricks workspace over two days — Unity Catalog scoping, the complete guardrail layer, deployed and yours to keep. This free session is the honest preview.

Talk about the private workshop

Can't make it live?

Register anyway — we'll send the replay within 24 hours so you can watch on your own schedule.

Register free — takes 10 seconds