RL Systems Architect — Scale Post-Training for LLMs

AItoolnavio; France; Remote; EUR 120,000 - 190,000; Full time

Job summary

Luma is seeking an expert to design, build, and scale distributed RL post-training systems for large-scale LLMs. You will orchestrate trainer, rollout, environment, and reward workloads across thousands of GPUs, with high-throughput rollout generation and asynchronous weight synchronization.

You will develop RL environments, reward infrastructure, and verifiers, plus evaluation tooling to keep runs stable. This role emphasizes scaling efficiency and rapid productionization with researchers.

Qualifications

• Hands-on experience post-training LLMs with RL at scale (PPO/GRPO/RLHF).
• Extensive distributed PyTorch training for foundation models.
• Experience building RL environments, reward functions, verifiers, or evaluation harnesses.
• Familiarity with RL post-training frameworks (veRL/OpenRLHF/TRL) and rollout engines (vLLM/SGLang).
• Understanding of GPU clusters, networking, and libraries (NCCL, MPI).

Responsibilities

• Design, build, and scale distributed RL post-training systems across thousands of GPUs.
• Build high-throughput rollout generation with inference engines and async weight syncing.
• Create RL environments for multi-step tasks with sandboxed execution and tool use.
• Develop verifiable rewards, LLM-as-judge pipelines, and defenses against reward hacking.
• Develop evaluation, monitoring, and debugging tooling for large RL runs.
• Turn new post-training ideas into production with researchers.

Skills

RL with LLMs

Distributed PyTorch training

RL environments & evaluators

RL frameworks

Rollout inference engines

GPU clusters & tooling

Tools

Kubernetes

Ray

NCCL

MPI

vLLM

SGLang

Job description

Luma is seeking an expert to design, build, and scale distributed RL post-training systems for large-scale LLMs. You will orchestrate trainer, rollout, environment, and reward workloads across thousands of GPUs, with high-throughput rollout generation and asynchronous weight synchronization.

You will develop RL environments, reward infrastructure, and verifiers, plus evaluation tooling to keep runs stable. This role emphasizes scaling efficiency and rapid productionization with researchers.