Senior AI Inference Scientist: LLM & VLM Optimization

Nebius Group; Zürich; On-site; CHF 140,000 - 190,000; Full time

Benefits offered by this job

Competitive compensation

Career growth and learning机会

Flexibility and ownership

Collaborative and innovative culture

Opportunity to work on impactful AI项目

International environment and talented

Job summary

Nebius is seeking scientists who can turn frontier LLM and VLM inference bottlenecks into research questions, conduct rigorous experiments, and translate their findings into production capabilities.

You will own focused research on quantization, distillation, speculative decoding, KV-cache optimization, and model/runtime co-optimization, working with engineering teams to bring promising results into production.

Qualifications

• PhD in computer science, ML systems, or closely related field.
• Strong publication record or equivalent artifacts in ML/efficient inference.
• Proficiency in Python and PyTorch with ability to build experiments and prototypes.

Responsibilities

• Lead research projects in efficient LLM/VLM inference from hypotheses to production handoff.
• Develop methods spanning quantization, QAT, distillation, speculative decoding, KV-cache reuse/compression, long-context inference, MoE routing.
• Build prototypes using PyTorch, Triton, CUDA-adjacent tooling and collaborate with engineers to productionize results.
• Explore LLM request routing, cache-aware load balancing, and prefill–decode capacity allocation to optimize latency and cost.
• Study multi-node inference, WideEP, expert placement and trade-offs for scalable architectures.
• Provide rigorous evaluations on quality, latency, throughput, memory, and cost per token under consistent workloads.
• Collaborate with MLE, kernel/back-end teams, product, and customers to set priorities with measurable impact.
• Share results via internal reports, blogs, papers, and open-source artifacts and mentor others.

Skills

Python

PyTorch

Quantization

Distillation

Experiment design

Research methods

Education

PhD in CS/ML/ML Systems

Tools

Triton

CUDA

PyTorch internals

NVIDIA software

Job description

Nebius is seeking scientists who can turn frontier LLM and VLM inference bottlenecks into research questions, conduct rigorous experiments, and translate their findings into production capabilities.

You will own focused research on quantization, distillation, speculative decoding, KV-cache optimization, and model/runtime co-optimization, working with engineering teams to bring promising results into production.