From Transactions to Agents: PostgreSQL in Modern AI Applications
Presented by:
AMEY BANARSE
Amey Banarse is a senior Field Engineering leader at Databricks, where he partners with AI-first digital-native and HealthTech companies to design and scale data-intensive, AI-driven applications. With nearly two decades of experience across distributed databases, cloud-native architectures, and analytics platforms, he has led mission-critical initiatives at companies such as YugabyteDB, Pivotal, and FINRA, and driven large-scale data architecture transformations for organizations including Paramount+, Wells Fargo, and General Motors. Amey is a frequent industry speaker and thought leader, having presented at AWS re:Invent, KubeCon, and Postgres-focused conferences on topics spanning Distributed SQL, agentic AI systems, and large-scale data platforms.
No video of the event yet, sorry!
Modern AI applications are no longer just chatbots—they are agentic systems that reason, retrieve context, call tools, and maintain state across many user interactions. For application developers, the hardest part isn’t model selection—it’s building an architecture that scales without sacrificing latency, correctness, or developer velocity. In this session, we’ll walk through a PostgreSQL-centric architectural pattern for building scalable agentic AI applications using Databricks Lakebase (Postgres-compatible OLTP) alongside the Databricks Lakehouse.
You’ll learn how to design a Compound AI System where: The Lakehouse acts as the agent’s long-term memory (analytics, embeddings, historical knowledge) Lakebase (PostgreSQL) acts as the agent’s operational brain (sessions, state, feedback, and transactional workflows) We’ll show how application developers can use familiar PostgreSQL concepts—transactions, schemas, indexes, and extensions like pgvector—to solve real-world AI application challenges such as low-latency state management, multi-agent coordination, and real-time feedback loops.
The talk concludes with a real-world customer case study, demonstrating how this pattern is used in production to power a high-scale, low-latency agentic application—without managing Kubernetes clusters, custom ETL pipelines, or bespoke vector infrastructure.
What Attendees Will Learn How to design agentic AI applications using PostgreSQL as a first-class runtime component Where PostgreSQL fits in modern RAG and tool-calling architectures How to combine OLTP (Lakebase) and analytics + vector search (Lakehouse) without fragile pipelines Practical patterns for: Agent session management Short-term vs long-term memory Feedback loops and observability How to scale agentic applications without rewriting your app stack
Key Takeaway If you know how to build applications with PostgreSQL, you already have most of the skills needed to build production-grade agentic AI systems. This session shows how to apply those skills—at scale—using a modern Lakehouse architecture.
- Date:
- 2026 April 21 14:00 PDT
- Duration:
- 20 min
- Room:
- San Pedro (Level C)
- Conference:
- Postgres Conference: 2026
- Language:
- Track:
- Dev
- Difficulty:
- Medium