SGLang Summit 2026 agenda: RL across B300/H200/MI355X, LinkedIn's 10K QPS ad serving
DavidBennett__ · x · 2026-11-06
SGLang Summit 2026 runs Nov 12-13 at Fort Mason, San Francisco. Weyaxi will present RL on heterogeneous hardware: single runs across B300, H200, MI355X and MI300X nodes, with measured results and upstream PRs.
Agenda highlights:
- LinkedIn serves a small ads-ranking model at 10K+ QPS with sub-80ms scoring latency
- Meta brings PyTorch to SGLang; Alibaba extends HiCache for long-running agents
- Perplexity shows a minimalist trillion-parameter trainer built with coding agents; Modal shares agents managing SGLang inference services
- NVIDIA on SGLang for Vera Rubin; Google on SGLang-TorchTPU for global-scale production inference
- Decagon on RL cutting hallucinations in customer agents; SGLang Omni supports interruptible realtime conversations
Related event: SGLang Summit 2026 Agenda Revealed: Focus on RL and Inference(2 posts)→
More from coding & agent
- Wealthsimple confirms Claude agent integration coming early 2027 — RichardsonDx · 2026-11-06
- Dev finds 5 performance disasters in Claude-generated code, fixed by vercel-react-best-practices skill — julianharris · 2026-11-06
- Building a work-focused agent harness: chat isn't enough, you need a workspace — GGTech407 · 2026-11-06
- The Church of Approximate Grace: an MCP server that sends agents to weekly services — coleclifford · 2026-11-06
- A five-agent team ran a startup for a day: 1h46m total runtime, onboarding to payments covered — pramodk73 · 2026-11-06
- Users claim Claude performs better with an empty system prompt than Anthropic's official one — repligate · 2026-11-06