LLM Post-Training Engineer at EPFL and ETH Zurich (Apertus Open Foundation Models)

Job Description

This LLM post-training job at EPFL supports Apertus, the open foundation-model project run jointly by EPFL and ETH Zurich, building and running supervised fine-tuning, preference optimisation and reinforcement learning pipelines on the Alps supercomputer of the Swiss National Supercomputing Centre (CSCS).

At a glance

• Position: Post-training engineer, 80–100%, fixed term (EPFL reference 2192)
• Where: EPFL Lausanne or ETH Zurich, Switzerland; remote-work options offered
• Duration: One year, renewable; start as soon as possible
• Apply: open until filled (re-checked 24 October 2026)

What the LLM post-training engineer does

• Build and maintain containers for LLM post-training and RL workloads, adapted to the Alps system at CSCS
• Run and monitor Slurm training and evaluation jobs, and debug distributed runs, checkpointing, filesystem, network and GPU problems
• Support SFT, preference optimisation and RL workflows, including RL environments with verifiable outcomes in maths, code, tool use and reasoning
• Work on reward modelling, reward calibration and verifier-based training, and generate and check synthetic training tasks
• Run ablations and evaluate models on reasoning, coding, maths, instruction following, multilingual, tool-use and safety benchmarks

Why this role matters

Most post-training know-how sits inside a few companies. Apertus is a publicly funded Swiss effort to build open foundation models, so the recipes, code and results are meant to be shared. The work is hands-on: making RL for LLMs run reliably at scale on a national supercomputer, next to the researchers who design the experiments.

What EPFL is looking for

• An MSc or PhD in computer science, data science, AI, machine learning or a related field; exceptional BSc graduates with strong engineering experience are considered
• Experience with AI and neural network architectures
• Good collaboration and communication across research and engineering teams

Nice to have

• Slurm or another HPC workload manager, and building containers for GPU clusters
• LLM fine-tuning, preference optimisation or RL (for example RLVR and reward modelling)
• Distributed training concepts (data, tensor and pipeline parallelism) and tools such as veRL, Megatron-LM, DeepSpeed, TRL, vLLM or SGLang
• Lower-level libraries such as NCCL, Transformer Engine or FlashAttention, and large evaluation pipelines

Pay and location

The ad does not state a salary. The role is at 80–100%, based at EPFL in Lausanne or ETH Zurich, with flexible and remote-work options, conference attendance and training.

How to apply

Apply through the official EPFL job posting. EPFL’s posting does not give a closing date, so the role is open until filled; ResearchJobs.in will re-check this listing on 24 October 2026. Check the posting for location and work-authorisation details before you apply.

See all our artificial intelligence jobs, or browse more research jobs on ResearchJobs.in.