Latest AI JobsBrowse Jobs
Luma AI logo

Research Scientist / Engineer – Reinforcement Learning Infrastructure

Luma AILondon, UKResearch & AI
Apply on company site
Salary
Not listed
Job type
Full-time
Work mode
Remote
Experience
Senior
Posted
8 Oct 2026
Status
Open · checked 7d ago

About this role

Research Scientist / Engineer – Reinforcement Learning Infrastructure is a full-time position based in London, UK, focused on building distributed systems for large-scale reinforcement learning. This role suits infrastructure and machine learning professionals with experience operating reinforcement learning loops at scale across thousands of GPUs.

Role Overview

Luma AI seeks a Research Scientist / Engineer to build the systems that drive reinforcement learning at frontier scale in London. The professional in this position couples policy optimization with large fleets of inference workers, agentic environments, and verification systems to transform model behavior into learning signals.

Key Responsibilities

Design, build, and scale distributed reinforcement learning post-training systems, coordinating trainer, rollout, environment, and reward workloads across thousands of GPUs.
Construct high-throughput rollout generation by integrating inference engines, weight synchronization, and asynchronous or off-policy schemes.
Formulate reinforcement learning environments for agentic, multi-step tasks including sandboxed code execution, tool use, computer use, and multimodal interaction that scale to millions of episodes.
Develop reward infrastructure encompassing programmatic rewards, reward-model serving, LLM-as-judge pipelines, and defenses against reward hacking.
Create evaluation, monitoring, and debugging tools to maintain stability during large reinforcement learning runs.
Advance training efficiency and stability while moving new post-training concepts into production runs alongside researchers.

Requirements

Practical experience post-training large language models with reinforcement learning such as PPO, GRPO, RLHF, and RLVR at significant scale.
Extensive background in distributed PyTorch training and parallelism techniques including FSDP, Tensor, Pipeline, and Expert Parallel for foundation models.
Experience constructing reinforcement learning environments, reward functions, verifiers, or evaluation harnesses for language model agents, featuring sandboxed execution and multi-turn tool use.
Familiarity with reinforcement learning post-training frameworks like veRL, OpenRLHF, TRL, Ray orchestration, and rollout inference engines including vLLM and SGLang.
Understanding of GPU clusters, networking, and communication libraries such as NCCL and MPI under mixed training and inference workloads.

Nice to Have

Executing reinforcement learning training across one hundred or more GPUs, utilizing asynchronous or disaggregated trainer and rollout architectures.
Containerization and orchestration platforms like Kubernetes and Ray for large environment fleets and sandboxed workloads.
Research contributions involving reinforcement learning for large language models or open-source contributions to reinforcement learning training frameworks.

Skills and Tools

PyTorch
FSDP
Tensor, Pipeline, and Expert Parallel
veRL
OpenRLHF
TRL
Ray
vLLM
SGLang
NCCL
MPI
Kubernetes

About the Company

Luma AI strives to build unified general intelligence capable of generating, understanding, and operating within the physical world. The organization emphasizes multimodality as a vital component for intelligence, positioning vision as the subsequent evolution beyond language models. Luma operates as an equal opportunity employer.

How to Apply

Interested candidates should apply directly on the official Luma AI careers page.

Skills and tech stack

PyTorchFSDPTensor ParallelismPipeline ParallelismExpert ParallelismveRLOpenRLHFTRLRayvLLMSGLangNCCLMPIKubernetesLLMLLMsReinforcement Learning

Questions candidates ask

Where is this role located?

This position is located in London, UK.

What core technologies are used in this role?

The tech stack includes PyTorch, FSDP, Ray, vLLM, SGLang, veRL, OpenRLHF, TRL, NCCL, MPI, and Kubernetes.

What is the primary focus of the job?

The role focuses on building distributed systems and infrastructure for reinforcement learning post-training at frontier scale across thousands of GPUs.

Are there any nice-to-have qualifications listed?

Yes, preferred qualifications include running reinforcement learning training across 100+ GPUs with asynchronous architectures, experience with Kubernetes and Ray for orchestration, and prior research or open-source contributions in reinforcement learning for language models.

What is Luma AI's mission?

Luma AI's mission is to build unified general intelligence that can generate, understand, and operate in the physical world.

Ready to apply?

The application is on the official Luma AI careers page.

Apply on company site