Maroon Ayoub
where else to find them
what they talk about
where they keep going back
no series more than once — all 1 in the filters
on a stage with
- Hyunkyun Moononce
- Kellen Swainonce
- Martin Hickeyonce
all 4 in the filters
who talks about the same things
- Junchen Jiang16 talks
- Yihua Cheng5 talks
- Ziwen Ning2 talks
- Yahav Biran7 talks
- Rafael Vasquez7 talks
- Zhou Sun2 talks
| title | conference | date |
|---|---|---|
| Why WideEP Inference Needs Data-Parallel-Aware Scheduling | PyTorch Conference Europe 2026 | 2026-04-07 |
WideEP fails not from expert cost but because routing ignores where KV state lives. Using vLLM plus LMD and a router extension called llm-d that reads KV-cache… | ||
| The Token Slice: Implementing Preemptive Scheduling Via Chunked Decod... | PyTorch Conference Europe 2026 | 2026-04-07 |
An experimental Chunked Decoding system slices generation into fixed-size token time slices and runs as a sidecar implementation for PyTorch servers (example… | ||
| Lightning Talk: KV-Cache Centric Inference: Building a State-Aware... | PyTorch Conference Europe 2026 | 2026-04-07 |
KV-cache centric inference reframes serving so KV state becomes the primary resource, implemented with llm-d plus vLLM, LLMCache and Envoy on Kubernetes; it… | ||
| Lightning Talk: Not All Tokens Are Equal: Semantic KV-Cache for Agen... | PyTorch Conference Europe 2026 | 2026-04-07 |
A runtime called Semantic KV-Cache replaces anonymous KV blocks in llm-d and vLLM with typed state that tags tokens as SystemPrompt, ToolDefinition… | ||
| nothing matches | ||