Research Scientist Engineer Reinforcement Learning Infrastructure
Luma AI develops multimodal creative AI agents, video and image generation models, and an API for generative media workflows.
Maintainer signals as of 9/25/2026
Funding history
About Luma AI
Luma AI is an active AI company whose Luma Agents product plans, generates, iterates, and refines creative work across video, image, audio, and text. Its first-party materials describe proprietary Ray video and Uni image models, as well as developer API access.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will design and scale distributed reinforcement-learning post-training systems across thousands of GPUs. You will build rollout, environment, reward, evaluation, monitoring, and debugging infrastructure; integrate inference engines and asynchronous training schemes; and improve training efficiency and stability.
Requirements
- Hands-on experience post-training LLMs with RL at meaningful scale
- Extensive distributed PyTorch training and parallelism experience for foundation models
- Experience building RL environments, reward functions, verifiers, or evaluation harnesses for LLM agents
- Deep familiarity with veRL, OpenRLHF, TRL, Ray, vLLM, and SGLang
- Strong understanding of GPU clusters, networking, NCCL, and MPI under mixed training and inference workloads
Responsibilities
- Design, build, and scale distributed RL post-training systems across thousands of GPUs
- Build high-throughput rollout generation with inference engines, weight synchronization, and asynchronous schemes
- Design scalable RL environments for agentic multi-step tasks
- Build reward infrastructure, including verifiable rewards, reward-model serving, LLM-as-judge pipelines, and reward-hacking defenses
- Develop evaluation, monitoring, and debugging tools for large RL runs
- Improve training efficiency and stability and turn post-training ideas into production runs