AI performance engineering meetup: GPU rooflines, agent trace-to-training loop, Homa transport
cfregly · x · 2026-10-19
Chris Fregly and Antje Barth host an online AI Performance Engineering meetup (Oct 19, 9–10 AM PDT) with three technical talks:
- GPU utilization under concurrency: Rajat Dandekar (Vizuara co-founder, Purdue PhD) visually explains why serving more users changes GPU utilization, tying vLLM measurements on an H100 to the roofline model.
- Agent traces to agent training: Rustem Feyzkhanov (Snorkel AI EM) walks through turning production agent failures into reproducible simulations, verifiable rewards, and fine-tuning/RL improvements — and when to change prompts, tools, or the agent harness instead of weights.
- Homa transport protocol: John Ousterhout presents Homa's fast lane for small messages and its potential for AI networking.
Aimed at engineers working on GPU inference, agent learning loops, and datacenter networking.
More from coding & agent
- Qwen Code v0.25.3 Ships Managed-Agent Fixes and Web-Shell Tweaks — qwen-code-review-bot · 2026-10-19
- Dad secretly installs GitHub Actions runners on his kids' computers for free CI — tristanbob · 2026-10-19
- Amazon Paper: QA Scores Miss Agent Memory Failures by 16.8 Points, New FEB-30 Benchmark Fixes It — rohanpaul_ai · 2026-10-19
- How One Founder Used Grok Bot to Submit Backlinks to 4000+ Sites in a Single Day — sujingshen · 2026-10-19
- Netflix shares how it built a knowledge base for production AI agents — AxSaucedo · 2026-10-19
- Reverse Engineer Anything crosses 300k GitHub stars as a movement — FinanceYF5 · 2026-10-19