Improving Cloud Reliability at Scale Using Gen AI
summary
Describes a deployed framework that uses GPT‑4–based agents to automate incident diagnosis by integrating telemetry, runbooks and debugging tools into a unified knowledge space. Built AIOpsLab and an agent called Flash agent that uses a hindsight corpus, tool invocation and an auto‑evaluation loop; trained on a corpus of 100,000 incidents, deployed to over 100 teams, and yields about 50% accuracy on recommendation tasks while enabling continuous improvement.
AI-written summary · about
chapters (10)
- 0:00Cloud reliability challenges
- 1:29Cloud reliability challenges
- 3:57Gen AI for cloud reliability
- 5:00LLM limitations in AIOps
- 6:05AIOps Lab for tools
- 7:52Agentic AI for incident diagnosis
- 9:17Handling AI hallucinations
- 10:40Flash in production incidents
- 11:25AIOps Lab framework
- 13:31Q&A
related talks
09mAI-Powered Cloud Automation – transforming cloud operations
17mAI TechOps: Data and Infra for AI Foundation to Scale and Manage 1000s of Cohesive Agents
30mAI Agents for Incident Management
31mTesting, Evaluating and Debugging Generative AI and Agentic Ap...
24mFrom First Experiments to AI Agents: Adoption Experience and Future Challenges
05mAutomating Cloud-Native Infra with Intelligent Tools
27mAI-Powered Reliability: Stop Fighting Fires, Prevent Them
11mAI-Native Network Operation