Open-Source Inference Engineering for the Agentic Era

Workshops Day 1Sponsor Session

Zain Hasan, Yubo Wang, Qingyang Wu, Jue Wang

When: Monday, June 299:00 AM – 11:00 AM · 120 min
Where: Track 8San Francisco, CA · imported from ai.engineer's public schedule feed

About this session

Agentic coding workloads demand long contexts, multi-turn conversations, and throughput at a scale that most inference engines weren't built for. TokenSpeed is a new open-source engine purpose-built for this regime, built collaboratively by NVIDIA DevTech, AMD Triton, Qwen Inference, Together AI, and others. In this 2-hour hands-on workshop, Together Inference Research Engineers and a TokenSpeed co-creator will cover TokenSpeed architecture, deploying your first model, optimizing for agentic workloads, kernel and hardware tuning, and throughput/latency trade-offs.

Speakers (4)

Zain Hasan

Staff AI/ML Engineer - DX, Together AI

AI/ML engineer and educator focused on large-scale models, tooling, and developer education.

Yubo Wang

LLM Inference, Together AI

Yubo Wang works on LLM inference at Together AI. His work and WF26 session focus on open-source inference engineering, serving large models efficiently, and building production inference systems for agentic workloads.

Qingyang Wu

Staff Research Scientist, Together AI

Qingyang Wu is a Staff Research Scientist at Together AI working on text generation, dialog systems, multimodal models and inference research. He previously researched language models at Columbia University.

Jue Wang

Senior Staff Researcher, Together AI

Jue Wang is a Senior Staff Researcher at Together AI working on efficient and cost-effective algorithms and systems for LLMs, after earning a Ph.D. in computer science from Zhejiang University.

More in Workshops Day 1

For developers: this programme is open data — JSON, iCal, schedule XML and an MCP endpoint.Show endpoints

No key, no signup, CORS open. Everything here is generated from the same data the organisers edit.