Senior Applied Scientist, Efficient LLM Inference & Model Optimization

Start free →

Or apply yourself on the official page →

About the role

About Nebius: 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. 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. Listed on Nasdaq (NBIS) and headquartered in Amsterdam, we have a global footprint with R&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&D. The role Nebius Token Factory is looking for 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 and optimization projects, write strong code, and produce prototypes, publications, and technical artifacts that engineers can build on. Your research will span quantization, distillation, speculative decoding, KV-cache optimization, and model/runtime co-optimization, alongside inference engines and distributed inference architectures. You will investigate how these approaches interact, evaluate their trade-offs across model quality, latency, throughput, memory footprint, and cost per token, and work with engineering teams to bring promising results into production. Your responsibilities • Lead research projects in efficient LLM and VLM inference, from hypotheses and experiments through ablations, prototypes, and production handoff. • Develop and evaluate methods spanning quantization, quantization-aware training (QAT), distillation, speculative decoding, KV-cache reuse, KV-cache compression, long-context inference, MoE routing, and model/runtime co-optimization. • Build prototypes using PyTorch, Triton, CUDA-adjacent tooling, or inference-serving frameworks, and collaborate with MLEs and platform engineers to turn them into production components. • Research LLM request routing and scheduling strategies, including cache-aware load balancing and prefill–decode disaggregation (PDD). Evaluate how queueing, KV-cache transfer, worker placement, and prefill/decode capacity allocation affect latency, throughput, and serving cost. • Investigate multi-node inference for dense and mixture-of-experts models, including wide expert parallelism (WideEP). Study expert placement, load imbalance, and computation/communication trade-offs, and use the findings to guide model and system architecture choices. • Develop rigorous evaluations covering quality, latency, throughput, numerical stability, memory footprint, tail latency, and cost per token. Compare serving architectures under consistent workload conditions and GPU budgets. • Work with MLE, GPU kernel, backend infrastructure, product, and customer teams to select research priorities with measurable production impact. • Share results through internal reports, technical blogs, papers, and open-source artifacts, and mentor engineers and scientists on experimental design, scientific rigor, and model/system trade-offs. Must-haves • A PhD in computer science, machine learning, ML systems, computer systems, computer architecture, electrical engineering, applied mathematics, or a closely related discipline. • A strong publication record or equivalent research artifacts in ML, ML systems, efficient inference, model compression, quantization, distillation, serving systems, or related areas. • Strong Python and PyTorch implementation skills, with the ability to turn ideas into experiments and wor

View the official posting →