LLM Post-Training & RL Engineer — Remote & HPC

ETH Zürich; Zürich; Hybrid; CHF 110.000 - 150.000; Vollzeit

Benefits dieser Stelle

Remote work options

Professional development

Open‑source projects

Zusammenfassung

ETH Zürich seeks an engineer to join the Apertus post-training effort, developing, running, and evaluating scalable post-training workflows for LLMs and reinforcement learning. The role emphasizes collaboration with researchers and CSCS, and work across Alps infrastructure in a fast-moving, research-driven HPC environment.

You will build containerised environments for post-training, adapt them for CSCS, and run Slurm-based training and evaluation jobs, contributing to open‑source, multilingual

Qualifikationen

• MSc or PhD in Computer Science, Data Science, AI, ML or related field.
• Hands-on experience with AI and neural network architectures.
• Experience with LLM post-training (SFT or RL).
• Strong collaboration across research and engineering teams.
• Hands-on experience with post-training workflows and RL environments.

Aufgaben

• Develop, run, and evaluate scalable post-training workflows for Apertus.
• Build and maintain containerised environments for LLM post-training and RL workloads.
• Adapt containers for Alps/CSCS infrastructure and run Slurm-based jobs.
• Debug distributed execution, checkpointing, filesystem performance, networking, and GPU utilisation.
• Collaborate with researchers and CSCS engineers to improve reliability of large-scale experiments.
• Support SFT, reinforcement learning workflows, reward modelling, and verifier-based training.
• Generate and validate synthetic or gym training tasks.
• Run ablation studies across algorithms, data mixes, hyperparameters, and infrastructure settings.
• Evaluate model behaviour across reasoning, coding, mathematics, instruction-following, multilingual, tool-use, and safety benchmarks.

Kenntnisse

LLM post-training

Software engineering

HPC collaboration

Python

Distributed training

Research communication

Ausbildung

MSc or PhD in CS/AI

BSc with strong engineering experience

Tools

Slurm

Containerization

GPU clusters

NCCL/DeepSpeed/Megatron-LM/TRL/vLLM

Jobbeschreibung

ETH Zürich seeks an engineer to join the Apertus post-training effort, developing, running, and evaluating scalable post-training workflows for LLMs and reinforcement learning. The role emphasizes collaboration with researchers and CSCS, and work across Alps infrastructure in a fast-moving, research-driven HPC environment.

You will build containerised environments for post-training, adapt them for CSCS, and run Slurm-based training and evaluation jobs, contributing to open‑source, multilingual