Senior Inference Engineer

Responsibilities: — Own the inference serving stack including prefill/decode disaggregation, continuous batching, KV-cache management, speculative decoding, MoE serving, quantization, and GPU performance engineering using CUDA and Triton. — Set the serving and measurement roadmap while remaining hands-on. Requirements: — Strong production experience with large language model (LLM) inference systems such as vLLM, SGLang, TensorRT-LLM or similar. — Deep knowledge of GPU performance including FP8/FP4 precision formats, memory bandwidth, and FLOPs. — Experience with Mixture of Experts (MoE) serving. — Rigorous benchmarking skills. — Technical leadership capabilities. — English proficiency at C1 level. — Open-source contributions to inference frameworks are a strong plus. Benefits: — Real ownership of projects. — Small senior team with low bureaucracy. — Competitive salary with performance-based incentives. — 24 days of annual leave plus public holidays. — Health insurance.