Free Live Webinar Series

Agentic Memory on the Lakehouse — built live, what actually breaks

A live-build series from the creator of the open-source db-agent. No slides — real code, real failures, real fixes. One registration covers the series.

Next session

Agentic Memory on Databricks: Deploy DB Agent with Redaction-First Memory

Monday, October 12 · 5:00 – 6:00 PM EDT

Reserve your free seat

Microsoft Teams link sent instantly. Replay + the session notebook sent to all registrants.

Free · No credit card · Teams link sent instantly

Monday, October 12 · 5:00 – 6:00 PM EDT
45–60 min · live build + Q&A
Free · notebook included
In this session
Live deploy: DB Agent fully on Databricks — Lakebase as the database, no separate vector service needed
How Lakebase does double duty: the same Postgres-compatible instance serves as both the OLTP data source and, via the pgvector extension it ships natively, the vector store for agentic memory
What “agentic memory” actually means here — and why session-recall memory (most of what's marketed as agentic memory) doesn't survive contact with enterprise security boundaries
The summarize-then-store pattern: what gets redacted, what gets kept, and why the redaction step is a second LLM call, not a filter
Live demo: two agents, two platforms, one shared memory store on Lakebase — zero raw data crossing the boundary
Why pgvector on Lakebase beats a bolted-on vector service for this: similarity ranking, TTL filtering, and self-exclusion in one indexed Postgres query, no client-side over-fetching
The exact bug this architecture caught in production testing (a silent embedding truncation) and what the fix looked like
Live Q&A — bring the version your security team already said no to
C

Chandan Kumar

Founder, beCloudReady · Creator of open-source db-agent (AAAI-25 workshop) · Organizer, TorontoAI (10K+ members)

The series

Each session is built from features actually shipped in db-agent — and the failures found shipping them. Register once; we'll notify you as each session goes live.

01Fully on Databricks: Lakebase as database and vector store, redaction-first agent memory via pgvectorNext session
02AWS S3 Vectors as an alternative shared memory store, for teams off DatabricksComing up
03The silent embedding bug that broke cross-agent memoryComing up
04Knowledge files: why schema alone makes agents hallucinateComing up
05What Databricks teaches vs. what we shipped — Genie, compared honestlyComing up

Who this series is for

Data & analytics engineers

You've built or evaluated a text-to-SQL agent and hit the wall every demo hides: real warehouses, real security zones, real business vocabulary the schema doesn't carry.

Platform & security leads

Your team wants agent memory; you can see the data leak. This session shows the architecture that separates shared context from shared data — with the redaction step explicit and auditable.

Engineers building on db-agent

You found the repo (probably via Google). These sessions are the maintainer walking through the newest shipped features — Lakebase as database and vector store, pgvector memory, knowledge files — with the reasoning behind each.

Data leaders evaluating agentic AI

You need to know what production agents actually break on before committing a team to one. Every session is built from shipped open-source code and real failures, not vendor decks.

#1 on Google for “db agent”AAAI-25 Workshop ProjectOpen Source on GitHubDatabricks Registered Partner

For data teams

Want this built with your team, on your data?

We run private, hands-on workshops for data teams — AI agents on Databricks, AI/BI Genie, Microsoft Copilot, and analytics workflows with AI tools — on your own workspace, with your own governance. The free series is the honest preview.

Talk about a private team workshop

Can't make it live?

Register anyway — every registrant gets the replay and the session notebook within 24 hours.

Register free — takes 10 seconds