Senior Machine Learning Engineer, LLM Inference Optimization

div class= content-intro p strong About Nebius: /strong /p p Nebius is leading a new era in cloud infrastructure for the global AI economy. We are building a full-stack AI cloud platform that supports developers and enterprises from data and model training through to production deployment, without the cost and complexity of building large in-house AI/ML infrastructure. /p p Built by engineers, for engineers. From large-scale GPU orchestration to inference optimization, we own the hard problems across compute, storage, networking and applied AI. /p p Listed on Nasdaq (NBIS) and headquartered in Amsterdam, we have a global footprint with R amp;D hubs across Europe, the UK, North America and Israel. Our team of 1,500+ includes hundreds of engineers with deep expertise across hardware, software and AI R amp;D. /p /div h2 The role /h2 p As a Senior Machine Learning Engineer on nbsp;the Applied AI team at Nebius Token Factory, nbsp;you will own model and endpoint optimization from model artifacts through production deployment. Your work will span model internals, nbsp;quantization and model compression, speculative decoding, KV-cache optimization, nbsp;inference engines, serving architecture, and nbsp;benchmarking to improve nbsp;latency, throughput, memory efficiency, GPU utilization, and cost per token while nbsp;preserving nbsp;quality and reliability. /p p You nbsp;will work with kernel and platform nbsp;engineers to investigate serving problems, compare configurations, and nbsp;resolve performance and quality nbsp;regressions. You will evaluate model- and engine-level optimizations together with distributed inference designs, including request routing, scheduling, prefill/decode coordination, and multi-node GPU execution. Your improvements will be validated through nbsp;reproducible benchmarks and safe production nbsp;rollouts under real-world workloads. /p h2 Your responsibilities /h2 ul li Own optimization nbsp;projects nbsp;for specific model families, customer endpoints, or serving backends. /li li Compare inference engines nbsp;and recommend serving configurations nbsp;suited to individual nbsp;workloads. /li li Diagnose model-quality and nbsp;performance regressions during production rollouts. /li li Improve nbsp;LLM and VLM nbsp;endpoint nbsp;latency, throughput, memory efficiency, GPU utilization, quality, and cost per token. /li li Deploy, configure, benchmark, and extend engines such as vLLM, SGLang, TensorRT-LLM, Triton Inference Server, nbsp;and NVIDIA Dynamo. /li li Develop production model-compression workflows covering nbsp;quantization, quantization-aware training, distillation, low-bit serving, and accuracy recovery. /li li Implement or integrate speculative decoding, draft-model approaches, KV-cache optimization, prefix caching, chunked prefill, continuous batching, and disaggregated prefill/decode serving. nbsp;For prefill–decode disaggregation (PDD), evaluate KV-cache transfer, worker placement, and capacity balancing to determine which workloads benefit from the architecture. /li li Design and improve LLM request routers and scheduling policies that balance worker utilization, queueing, request characteristics, and KV-cache locality while meeting latency and reliability targets. /li li Scale dense and mixture-of-experts inference across multiple GPU nodes. Select and tune tensor, pipeline, data, and expert parallelism—including wide expert parallelism (WideEP)—with attention to hardware topology, expert load balance, and communication overhead. /li li Build reproducible benchmark harnesses nbsp;measuring nbsp;TTFT, TPOT, tokens per second per GPU, p95/p99 latency, GPU memory, reliability, and cost per token. nbsp;Validate architecture choices under representative traffic and consistent GPU budgets. /li li Collaborate nbsp;with GPU kernel and platform engineers to nbsp;trace nbsp;bottlenecks across model code, kernels, runtime, scheduler, gateway, and cluster layers. /li li Produce design documents, nbsp;performance reports, rollout plans, and customer-facing technical explanations. /li /ul h2 Must-haves /h2 ul li Strong nbsp;engineering skills in Python and PyTorch. /li li Hands-on experience deploying or optimizing LLM, VLM, or high-throughput transformer nbsp;inference. /li li Practical knowledge of at least one modern inference nbsp;stack, nbsp;such as vLLM, SGLang, TensorRT-LLM, Triton Inference Server, NVIDIA Dynamo, Ray Serve, KServe, or nbsp;an equivalent internal system. /li li A strong nbsp;understanding of transformer inference nbsp;bottlenecks involving nbsp;KV cache, attention, memory bandwidth, batching, parallelism, and long-context serving. /li li Ability to nbsp;quantify trade-offs among nbsp;latency, throughput, quality, utilization, and nbsp;cost. /li li Experience designing or optimizing distributed inference system architecture, with hands-on work in one or more areas such as request routing, distributed scheduling, PDD, multi-node inference, or expert parallelism. /li li Strong communication and nbsp;collaboration skills across nbsp;research, kernel, infrastructure, product, and customer teams. /li /ul h2 Nice-to-haves /h2 ul li Experience with quantization-aware training, post-training quantization, FP8, INT8, INT4, NVFP4, MXFP4, AWQ, GPTQ, or nbsp;SmoothQuant. /li li Experience with distillation, speculative decoding, EAGLE, Medusa, multi-token prediction, or nbsp;related nbsp;inference acceleration methods. /li li Experience nbsp;supporting agentic workloads involving nbsp;tool calling, structured outputs, streaming APIs, high concurrency, and multi-step orchestration. /li li Familiarity with CUDA or Triton; the role does not require kernel engineering to be the candidate’s primary specialization. /li li Contributions to open-source projects such as nbsp;vLLM, SGLang, TensorRT-LLM, FlashInfer, LMCache, PyTorch, Triton, Ray, or nbsp;KServe. /li li Implementation experience with cache-aware request routing, PDD, or WideEP, including diagnosing communication bottlenecks and load imbalance. /li li Familiarity with GPU communication libraries and interconnects, such as NCCL, NVLink, InfiniBand, or RoCE. /li /ul div class= content-conclusion p strong Benefits amp; Perks: /strong /p ul li Competitive compensation /li li Career growth and learning opportunities /li li Flexibility and ownership /li li Collaborative and innovative culture /li li Opportunity to work on impactful AI projects /li li International environment and talented teams /li /ul p strong What's it like to work at Nebius: /strong /p p Fast moving nbsp;- Bold thinking nbsp;- Constant growth nbsp;- Meaningful impact nbsp;- Trust and real ownership nbsp;- Opportunity to shape the future of AI nbsp; /p p strong Equal Opportunity Statement: /strong /p p Nebius is an equal opportunity employer. We are committed to fostering an inclusive and diverse workplace and to providing equal employment opportunities in all aspects of employment. We do not discriminate on the basis of race, color, religion, sex (including pregnancy), national origin, ancestry, age, disability, genetic information, marital status, veteran status, sexual orientation, gender identity or expression, or any other characteristic protected by applicable law. /p p Applicants must be authorized to work in the country in which they apply and will be required to provide proof of employment eligibility as a condition of hire. nbsp; /p p If you need accommodations during the application process, please let us know. /p /div Find more English Speaking Jobs in Netherlands on Arbeitnow