Research Engineer / Scientist, Agentic RL Post-Training (all genders)
Core
End-to-end post-training of LLMs for agentic work using reinforcement learning on synthetic tasks across hundreds of GPUs.
Role type
Senior Research Engineer / Scientist (Agentic RL)
Builds
Production-ready agentic models deployed to European companies and public administration
Domain
AI / Machine Learning / Reinforcement Learning
Deliverable
production ML models
Required skills
Reinforcement Learning (PPO/GRPO, RLHF), PyTorch, Distributed Training (FSDP, DeepSpeed, Megatron), Experiment Design, Algorithm Development, System Debugging
Preferred skills
vLLM, SGLang, verl, OpenRLHF, TRL, Multi-node debugging, GPU performance optimization
Technologies
PyTorch, EuroHPC, Slurm, Kubernetes, NCCL
Responsibilities
Design and execute on-policy RL runs, optimize algorithms and data mixtures, debug training instability, build reproducible training infrastructure, ship models to customers
Seniority
Senior, hands-on IC
