Where you can live
Open to people living in Austria, Belgium, Bulgaria, Cyprus, Czechia, Germany, Denmark, Estonia, Spain, Finland, France, Greece, Croatia, Hungary, Ireland, Italy, Lithuania, Luxembourg, Latvia, Malta, Netherlands, Poland, Portugal, Romania, Sweden, Slovenia, Slovakia.
Listed location: Remote, EU
Taken from the location the company listed. Confirm in the posting before you apply.
What the job is
Builds and scales distributed reinforcement learning systems for LLM post-training, including rollout generation, RL environments, reward infrastructure, and monitoring tools.
What they ask for
- Hands-on experience post-training LLMs with RL (PPO/GRPO-family, RLHF, RLVR) at meaningful scale
- Extensive distributed PyTorch training and parallelism (FSDP, Tensor/Pipeline/Expert Parallel)
- Experience building RL environments, reward functions, verifiers, or evaluation harnesses for LLM agents
- Deep familiarity with RL post-training frameworks (veRL, OpenRLHF, TRL, Ray) and rollout engines (vLLM, SGLang)
- Strong understanding of GPU clusters, networking, and communication libraries (NCCL, MPI)
Nice to have
- Running RL training across 100+ GPUs, including asynchronous or disaggregated architectures
- Containerization and orchestration (Kubernetes, Ray) for environment fleets and sandboxed workloads
- Research contributions in RL for LLMs or open-source contributions to RL training frameworks
This is a summary written by software from the company's posting. The posting is the source.
More remote jobs at Luma AI