Senior Machine Learning Engineer, LLM Inference Optimization
About this role
Delivering fast, reliable, and cost-efficient inference services for frontier models requires solving complex bottlenecks in latency, throughput, memory efficiency, and GPU utilization. The Nebius Applied AI team addresses this challenge by building scalable infrastructure for the global AI economy, enabling developers and enterprises to deploy models smoothly.
As a Senior Machine Learning Engineer focused on LLM inference optimization, this person will own model and endpoint optimization from raw artifacts to production deployment. Daily work involves deploying and benchmarking inference engines, debugging performance regressions, and partnering with kernel and platform engineers to resolve deep technical constraints. Ideal candidates bring strong engineering foundations in Python and PyTorch alongside hands-on experience optimizing high-throughput transformer inference systems.
What you will do
- Own optimization work for specific model families, customer endpoints, or serving backends.
- Run engine comparisons and recommend practical serving configurations for specific workloads.
- Debug model quality or performance regressions during production rollouts.
- Optimize LLM and VLM endpoints for latency, throughput, memory efficiency, GPU utilization, quality, and cost per token.
- Deploy, configure, benchmark, and extend inference engines such as vLLM, SGLang, TensorRT-LLM, Triton Inference Server, and NVIDIA Dynamo.
- Build and productionize model-compression workflows including quantization, quantization-aware training, distillation, and accuracy recovery.
- Implement or integrate speculative decoding, draft-model approaches, KV-cache optimization, prefix caching, chunked prefill, and continuous batching.
- Build reproducible benchmark harnesses for time-to-first-token, time-per-output-token, tokens per second, latency percentiles, memory, and cost.
What you need
- Demonstrated engineering proficiency using Python and PyTorch.
- Hands-on background deploying or optimizing LLM, VLM, or high-throughput transformer inference systems.
- Practical knowledge of at least one modern inference stack such as vLLM, SGLang, TensorRT-LLM, Triton Inference Server, NVIDIA Dynamo, Ray Serve, or KServe.
- Strong understanding of transformer inference bottlenecks involving KV cache, attention, memory bandwidth, batching, parallelism, and long-context serving.
- Ability to reason quantitatively about latency, throughput, quality, utilization, and cost tradeoffs.
- Strong communication skills and ability to collaborate with research, kernel, infrastructure, product, and customer teams.
Nice to have
- Experience with quantization-aware training, post-training quantization, FP8, INT8, INT4, NVFP4, MXFP4, AWQ, GPTQ, SmoothQuant, or related techniques.
- Experience with distillation, speculative decoding, EAGLE, Medusa, multi-token prediction, or other inference acceleration methods.
- Experience with agentic workloads, including tool calling, structured outputs, streaming APIs, high concurrency, and multi-step orchestration.
- CUDA or Triton familiarity, even if the role is not primarily a kernel-engineering role.
- Open-source contributions to vLLM, SGLang, TensorRT-LLM, FlashInfer, LMCache, PyTorch, Triton, Ray, KServe, or related projects.
Benefits
- Competitive compensation
- Career growth and learning opportunities
- Flexibility and ownership
- Collaborative and innovative culture
- Opportunity to work on impactful AI projects
- International environment and talented teams
Skills and tech stack
Questions candidates ask
What is the team building at Nebius?
The Nebius Token Factory is building fast, reliable, and cost-efficient inference services for frontier models. As part of the Applied AI team, engineers own model and endpoint optimization from model artifacts through production deployment.
What tools and technologies will I work with?
You will deploy, configure, benchmark, and extend inference engines such as vLLM, SGLang, TensorRT-LLM, Triton Inference Server, NVIDIA Dynamo, or similar systems. The role also requires strong Python and PyTorch engineering skills.
What are the primary responsibilities for this role?
You will optimize LLM and VLM endpoints for latency, throughput, memory efficiency, GPU utilization, quality, and cost per token. Additionally, you will build and productionize model-compression workflows and partner with GPU kernel engineers to diagnose bottlenecks.
What is the work environment like?
Nebius offers an international environment and talented teams with flexibility and ownership, career growth, and learning opportunities. The culture emphasizes fast-moving, bold thinking, constant growth, meaningful impact, trust, and real ownership.