How We Debug 1000s of Databases with AI: Lessons from an AI-Assisted Database...
summary
An AI-assisted platform debugs thousands of MySQL and TiDB instances across 70+ regions and three clouds by consolidating metrics, logs and KB articles, synthesizing timelines and recommending remediations; a live demo performs a parameter rollback and verifies completion in about five minutes. Architecture uses a Storax central-first sharded architecture, fine-grained RBAC, an internal LLM proxy (Claude/GPT/Gemini) and a DSPy-like modules framework, with a Temporal approval workflow and tested safety gates that drive adoption and cut investigation time by up to 90%.
AI-written summary · about
chapters (9)
- 0:00Team and Databricks intro
- 1:44Why AI for massive-scale ops
- 3:33AI agent debugging demo
- 6:44Reducing ops toil by 90%
- 7:27Infrastructure as a product
- 10:16Safe AI integration foundation
- 12:44Scaling SRE expertise with AI
- 16:18Reliable AI in production
- 16:59Closing and careers
related talks
31mAutopilot With Guardrails Designing AI Assisted Database Operations Without Losing Control
32mManaging Databases in the Cloud in Broken
50mLeveraging AI as a PostgreSQL DBA
30mAI-Assisted Post-Mortem Analysis for PostgreSQL
33mEngineer More, Work Less: DBS’ Autonomous Database Journey
20mOpen Source Database Operations in the Age of AI
18mManaging Databases in the Cloud is BROKEN!
45mForget SQL! Create your own AI Data Analyst