Senior Applied Scientist, Efficient LLM Inference

On-site · Zurich (Switzerland) · Senior · Full-time

Want offers like this one?

Create your free PyHires profile: you will see how well you match each offer and get a weekly email with the ones that fit you.

Create my profile

Overview

Work mode
On-site
Location
Zurich (Switzerland)
Engagement
Full-time
Seniority
Senior
Languages
English
Open to candidates from
All of Europe (EU and non-EU)
Company
Nebius
Posted
Nov 20, 2026

About the role

Nebius builds a full-stack AI cloud platform. Nebius Token Factory is looking for scientists who turn frontier LLM and VLM inference bottlenecks into research questions, run rigorous experiments and bring the results into production. Your research will cover quantization, distillation, speculative decoding, KV-cache optimisation, model/runtime co-optimisation and distributed inference. What you'll do - Lead research projects in efficient LLM and VLM inference, from hypotheses to prototypes and production handoff - Develop methods for quantization and QAT, distillation, speculative decoding, KV-cache reuse and compression, long context and MoE routing - Build prototypes with PyTorch, Triton, CUDA-adjacent tools or inference-serving frameworks - Research request routing and scheduling, including cache-aware load balancing and prefill-decode disaggregation - Study multi-node inference for dense and MoE models, including wide expert parallelism - Build rigorous evaluations of quality, latency, throughput, memory and cost per token - Share results through reports, blogs, papers and open source, and mentor others What you bring - PhD in computer science, machine learning, ML systems or a closely related field - Strong publication record or equivalent research artifacts in efficient inference or related areas - Strong Python and PyTorch, turning ideas into experiments and prototypes - Deep knowledge of LLMs, VLMs, transformer inference, decoding, compression and serving trade-offs - Strong experimental design and clear communication Nice to have - First-author papers at venues such as NeurIPS, ICML, ICLR or MLSys - Production deployment of ML models or inference optimisations - vLLM, SGLang, TensorRT-LLM, FlashAttention, Triton, CUDA or PyTorch internals - Post-training (SFT, DPO, RLHF) applied to inference quality or efficiency What they offer - Competitive compensation - Career growth, flexibility and ownership - International environment working on impactful AI projects

Requirements

Based on:Python

Must-have

  • LLMEssential
  • PyTorchEssential

About Nebius

Nebius builds a full-stack AI cloud platform that supports developers and enterprises from data and model training through to production deployment, covering GPU orchestration, storage, networking and inference optimization. Listed on Nasdaq (NBIS) and headquartered in Amsterdam, it has R&D hubs across Europe, the UK, North America and Israel and a team of more than 1,500 people.

Stack

  • Python
  • LLM
  • PyTorch

Company

Nebius logo
Nebius
Switzerland
View company profile