Building a Self-Service Platform for Stateful AI Agents
summary
Claims a self-contained data agents pattern that bundles runtime, tool contracts, and an in-agent OLAP so queries run locally and latencies drop from seconds to single-digit milliseconds. Uses Knative serving (Cloud Run demo) for Knative scale-to-zero plus activator/autoscaler, and local CHDB/DuckDB choices (ClickHouse or DuckDB) for in-memory SQL. Describes three data delivery patterns: fetch-on-boot, sidecar volume, sharded fleet. Demo runs natural language price queries against London property data, mostly works with one minor result bug.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 4:17Problem statement and evolution
- 7:27Data agents and their types
- 11:19Self-contained data agents benefits
- 14:27Design goals for stateful platforms
- 16:13Serverless architecture for AI agents
- 19:01Knative scale to zero
- 21:46Live demo: self-contained agent
- 25:19Key takeaways
- 26:44Q&A
related talks
24mBuilding Your Intelligent Data Agent
19mBuilding Agents at Cloud Scale
34mOpen Source AI Agents on User-Owned Infra: K3s, MCP, & GPU Sharing in Practice
05m⚡ Can Stateful AI Agents Scale To Zero?
07mOwn your Agentic AI
46mFullstack AI: The Building Blocks for Production Ready AI-Powered Software
35mMulti-Layered Guardrails for Cloud Native AI: Enforcin...
17mAI TechOps: Data and Infra for AI Foundation to Scale and Manage 1000s of Cohesive Agents