Senior Applied Scientist, Efficient LLM Inference & Model Optimization
About this role
Nebius Token Factory is building a full-stack AI cloud platform that supports developers and enterprises from data and model training through to production deployment. The team focuses on solving difficult challenges in efficient LLM inference, model optimization, and GPU orchestration to remove friction in the global AI economy.
As a Senior Applied Scientist, you will turn frontier inference bottlenecks into research problems, publish credible work, and help ship the results into production. You will own focused research projects from hypothesis through experiment, ablation, prototype, and production handoff while partnering directly with machine learning engineers.
This position exists to invent, evaluate, and productionize methods for quantization, distillation, speculative decoding, and long-context inference with measurable impact. An ideal candidate combines a strong academic research background with hands-on coding ability in Python and PyTorch, bridging the gap between theoretical exploration and real-world deployment.
What you will do
- Own focused research projects spanning hypothesis formulation, experimentation, ablation, prototyping, and final production handoff.
- Define and execute research programs focused on efficient LLM and VLM inference with measurable impact on production systems.
- Invent, evaluate, and productionize advanced methods for quantization, QAT, distillation, speculative decoding, and KV-cache optimization.
- Build robust prototypes using PyTorch, Triton, CUDA-adjacent tooling, and modern inference-serving frameworks.
- Design rigorous evaluation methodologies to measure quality, latency, throughput, memory footprint, numerical stability, and cost per token.
- Prepare internal reports, technical blogs, and academic papers to share findings and build external credibility for the organization.
- Collaborate across machine learning engineering, GPU kernel, backend infrastructure, product, and customer teams to select high-leverage research bets.
- Mentor engineers and scientists regarding experimental design, scientific rigor, and hardware-software system tradeoffs.
What you need
- Hold a PhD in computer science, machine learning, ML systems, computer systems, computer architecture, electrical engineering, applied math, or a closely related field.
- Demonstrate a strong publication record or equivalent research artifacts in machine learning, efficient inference, model compression, or related areas.
- Possess strong hands-on coding ability in Python and PyTorch to move quickly from ideas to working experiments and prototypes.
- Bring a deep understanding of large language models, vision-language models, transformer inference, decoding algorithms, and production-serving tradeoffs.
- Show strong experimental design skills including ablations, baselines, metrics, statistical reasoning, and failure analysis.
- Communicate effectively with excellent written and verbal skills.
Nice to have
- First-author publications in NeurIPS, ICML, ICLR, MLSys, ACL, EMNLP, ASPLOS, OSDI, SOSP, ISCA, HPCA, or comparable venues.
- Experience deploying machine learning models or inference optimizations directly in production environments.
- Familiarity with vLLM, SGLang, TensorRT-LLM, NVIDIA Dynamo, FlashAttention, FlashInfer, Triton, CUDA, or PyTorch internals.
- Experience with post-training, SFT, DPO, RLHF, RLAIF, preference optimization, or synthetic data generation connected to inference quality or efficiency.
- Open-source research artifacts, widely used benchmarks, high-quality technical blogs, or invited talks in efficient AI systems.
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 locations are available for this role?
The role is located across several hubs including Amsterdam, Berlin, London, Poland, Prague, Tel Aviv, and Zurich. Applicants must be authorized to work in the country in which they apply.
Is this strictly a research position or will I write production code?
This is not a papers-only research role. You are expected to design rigorous experiments, write strong code, collaborate with engineers, and convert research into deployed inference capabilities.
What specific optimization areas will I work on?
You will invent, evaluate, and productionize methods for quantization, QAT, distillation, speculative decoding, KV-cache reuse, KV-cache compression, long-context inference, MoE routing, and model/runtime co-optimization.
Who will I collaborate with on a daily basis?
You will partner directly with machine learning engineers, GPU kernel developers, backend infrastructure teams, product units, and customer teams to choose high-leverage research bets.