F

Senior Applied Scientist, Efficient LLM Inference & Model Optimization

FDE Times·Zürich·05.11.2026

 fdetimes.net

 
 
full_time80–100%
Job written in
English
Location
Zürich
Work type
On-site
Type
Full-time

 

Nebius is a cloud‑infrastructure company that powers the global AI economy with a full‑stack platform covering everything from data handling and model training to production deployment. Headquartered in Amsterdam and listed on Nasdaq, the firm operates R&D hubs across Europe, the UK, North America and Israel, employing more than 1,500 engineers who tackle large‑scale GPU orchestration, inference optimisation and other hard problems in compute, storage, networking and applied AI. The Senior Applied Scientist will focus on making large language model (LLM) and vision‑language model (VLM) inference more efficient. Day‑to‑day work involves leading research projects that start with hypotheses, move through rigorous experiments and ablations, and end with prototypes ready for production hand‑off. The role requires building and evaluating techniques such as quantisation, quantisation‑aware training, distillation, speculative decoding, KV‑cache reuse and compression, long‑context inference, mixture‑of‑experts routing and co‑optimisation of models with runtimes. Prototypes are created with PyTorch, Triton, CUDA‑adjacent tools or other serving frameworks, and the scientist collaborates closely with machine‑learning engineers, GPU‑kernel teams and platform engineers to integrate successful methods into Nebius's inference stack. A strong research background is essential. Candidates must hold a PhD in a relevant discipline (computer science, machine learning, ML systems, computer architecture, electrical engineering, applied mathematics or a closely related field) and demonstrate a solid publication record or equivalent research artefacts in areas such as efficient inference, model compression or serving systems. Technical expertise includes advanced Python and PyTorch coding, deep understanding of transformer‑based LLMs and VLMs, decoding algorithms and the trade‑offs involved in production serving. The scientist must also be skilled in experimental design, including ablations, baseline selection, metric definition, statistical analysis and failure investigation, and possess excellent written and verbal communication abilities. Preferred qualifications include first‑author papers at top conferences (NeurIPS, ICML, ICLR, MLSys, ACL, EMNLP, ASPLOS, OSDI, SOSP, ISCA, HPCA) and hands‑on experience deploying ML models or inference optimisations in production environments. Familiarity with tools such as vLLM, SGLang, TensorRT‑LLM, NVIDIA Dynamo, FlashAttention, FlashInfer, Triton, CUDA or internal PyTorch components is valued, as is experience with post‑training techniques (SFT, DPO, RLHF, RLAIF) that improve inference quality or efficiency. Contributions to open‑source research, widely used benchmarks, technical blogs or invited talks, as well as work on distributed inference architectures (request routing, prefill‑decode disaggregation, multi‑node serving, MoE expert parallelism) are also considered advantageous. Nebius offers a competitive compensation package, strong career‑growth opportunities, a flexible and ownership‑driven environment, and a collaborative, innovative culture. Employees work in an international setting with talented teams, tackling high‑impact AI projects that shape the future of the industry. The company emphasizes fast‑moving, bold thinking, constant growth and meaningful impact while maintaining an inclusive, equal‑opportunity workplace. What the role asks for: - PhD in computer science, ML, systems, architecture, EE, applied math or related field - Strong publication record or equivalent research artefacts - Advanced Python and PyTorch implementation skills - Deep knowledge of LLMs, VLMs, transformer inference and model compression - Expertise in experimental design, ablations and statistical analysis - Excellent written and verbal communication - First‑author papers at top AI/ML conferences (nice‑to‑have) - Experience with production deployment of ML inference optimisations (nice‑to‑have)

 

 

 

 

 

Kostenlos

Scientist: Lebenslauf-Vorlage

Bewirb dich mit einem Lebenslauf im Schweizer Aufbau — mit Beispieltext und den Anforderungen, die Scientist-Inserate am häufigsten nennen.

Lebenslauf-Vorlage Schweiz ansehen

Scientist: was der Markt gerade verlangt

Scientist-Stellen gehören zu den regelmässig ausgeschriebenen Berufen auf SwissJobs.app.

Weitersuchen

Ähnliche Jobs per E-Mail erhalten