vLLM Office Hours Hits 59 Sessions: 76 Speakers from 24 Organizations and Counting
vllm_project · x · 2026-11-07
The vLLM project recapped its bi-weekly Office Hours livestream run with Red Hat: 59 sessions so far, featuring 76 guest speakers from 24 organizations, with roughly 56 hours of recordings archived on YouTube.
Each 1-hour session has three parts:
- Project update (15 min): a core committer walks through recent releases
- Guest talk (30 min): a deep dive into a feature, hardware backend, or ecosystem project
- Open Q&A (15 min): deployment and contribution questions
The latest session (#58) introduced llm-d's semantic classifier (llm-d-sc), which picks a request's model tier before KV-cache routing selects a pod, including a comparison with vLLM Semantic Router and a NASA domain-routing proof of concept. Past topics span quantization (11 sessions), distributed inference (10), architecture, routing, and speculative decoding. The event is open to everyone via calendar subscription.
More from coding & agent
- Power User Spending 12-16 Hours Daily Argues Creativity, Not Models, Is the Bottleneck — tengyanAI · 2026-11-07
- Hugging Face Turns Reasoning Models into Agents with RL and LoRA Adapters — yb2698 · 2026-11-07
- Doki lets you make videos with no timeline — just write the whole thing as one story — thesephist · 2026-11-07
- UniGFM-RAG paper trains the graph retriever and LLM end-to-end for GraphRAG — JeremyCMorgan · 2026-11-07
- Dev controls coding agent via Rokid smart glasses to ship Black Friday campaign — yihui_indie · 2026-11-07
- Claude auto-renders 21k Blender images to train a model that recreates your LEGO build — nptacek · 2026-11-07