Latency & Throughput AI Research Engineer (LLM Inference)

Mistral is seeking a Research Engineer for the Latency & Throughput team, based in Paris or London with a hybrid setup. You will push the frontier of LLM inference by prototyping techniques to reduce latency and increase throughput, from speculative decoding to KV compression and new attention variants, bridging research and production.

This role sits between inference engineers and researchers. You will generate ideas, run experiments, and ship proof-of-concepts into our serving stack,