Description review
Research Scientist / Engineer – Reinforcement Learning Infrastructure
Luma AI · London, UK, Remote, UK · back to the listing
HR standards
48/100
poor
Title ↔ description
81/100
solid
Reads as
Machine Learning Engineer
98% confident
What this role officially is
data scientist — ESCO, the EU occupation classification
Data scientists find and interpret rich data sources, manage large amounts of data, merge data sources, ensure consistency of data-sets, and create visualisations to aid in understanding data. They build mathematical models using data, present and communicate data insights and findings to specialists and scientists in their team and if required, to a non-expert audience, and recommend ways to apply the data.
Also known as: data scientists, data engineer, research data scientist, data expert, data research scientist
How others title the same work
Large employers
- Applied AI Engineer Automattic
- Machine Learning Engineer, CX Intelligence Coinbase
- AI Engineer - FDE (Forward Deployed Engineer) Databricks
- AI Engineer - FDE (Forward Deployed Engineer) - U.S. Federal Sector Databricks
- Senior AI Engineer – Notebooks Datadog
Startups
- Attendi | Medior Machine Learning Engineer | Amsterdam, Netherlands | ONSITE (hybrid) | €6,000 - €7,000 per month | Full-time (80–100%, ~4–5 days/week) | Visa sponsorship + 30% ruling possible Attendi
- Member of Technical Staff (applied) Anthrogen
- Aptura AI | Full-Time | MTS (Applied AI), MTS (SWE / Product) | London | ONSITE / HYBRID Aptura AI
- Arcforma AI (arcforma.ai) | AI Engineer (Marketing / Construction / Arcforma AI (arcforma.ai)
- Staff AI Engineer - Agent Architecture & Behavior Artisan
What the listing never says
- 17 bullet points. Long requirement lists deter qualified candidates, who read them as hard gates. Scope clarity
- No section describes what the person would actually do. Scope clarity
- No pay range published. Candidates cannot tell whether applying is worth their time. Pay transparency
The listing, marked up
Nothing in the wording of this listing tripped a check. The scores above still judge how complete and coherent it is.
You'll build the systems that make reinforcement learning work at frontier scale — coupling policy optimization with large fleets of inference workers, agentic environments, and the reward and verification systems that turn model behavior into learning signal. RL is how Luma's models go from capable to useful.
RL at scale is a full-loop systems problem: training, rollout generation, environment execution, and reward computation running concurrently across thousands of GPUs, all needing to stay fast, stable, and correct together. It fits someone who has lived this — post-trained LLMs with RL, built environments and verifiers, and debugged asynchronous rollout pipelines at scale. If you haven't operated RL at real scale, this will be deep water.
What You'll Own
• Design, build, and scale distributed RL post-training systems, orchestrating trainer, rollout, environment, and reward workloads across thousands of GPUs.
• Build high-throughput rollout generation, integrating inference engines (vLLM, SGLang), weight synchronization, and asynchronous/off-policy schemes.
• Design RL environments for agentic, multi-step tasks — sandboxed code execution, tool use, computer use, multimodal interaction — reproducible and scalable to millions of episodes.
• Build reward infrastructure: verifiable/programmatic rewards, reward-model serving, LLM-as-judge pipelines, and defenses against reward hacking.
• Develop the evaluation, monitoring, and debugging tooling that keeps large RL runs stable.
• Advance training efficiency and stability, and turn new post-training ideas into production runs with researchers.
First 90 Days
One way the first 90 could unfold.
• Days 1–30 — Immerse & Diagnose: Learn the current RL stack and where throughput, stability, or correctness break.
• Days 30–60 — Ship & Validate: Improve a piece of the loop (rollout throughput, reward infra, or an environment) and prove it on a real run.
• Days 60–90 — Scale & Systemize: Harden the full loop across thousands of GPUs and asynchronous architectures.
What You Bring
• 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) for foundation models.
• Experience building RL environments, reward functions, verifiers, or evaluation harnesses for LLM agents, including sandboxed execution and multi-turn tool use.
• Deep familiarity with RL post-training frameworks (veRL, OpenRLHF, TRL, Ray orchestration) and rollout inference engines (vLLM, SGLang).
• Strong understanding of GPU clusters, networking, and communication libraries (NCCL, MPI) under mixed training and inference workloads.
Nice to Have
• Running RL training across 100+ GPUs, including asynchronous or disaggregated trainer/rollout architectures.
• Containerization and orchestration (Kubernetes, Ray) for large environment fleets and sandboxed workloads.
• Research contributions in RL for LLMs, or open-source contributions to RL training frameworks.
About Luma: Luma's mission is to build unified general intelligence that can generate, understand, and operate in the physical world. We believe multimodality is critical for intelligence — the next step beyond language models comes from vision. Luma is an equal opportunity employer.
RL at scale is a full-loop systems problem: training, rollout generation, environment execution, and reward computation running concurrently across thousands of GPUs, all needing to stay fast, stable, and correct together. It fits someone who has lived this — post-trained LLMs with RL, built environments and verifiers, and debugged asynchronous rollout pipelines at scale. If you haven't operated RL at real scale, this will be deep water.
What You'll Own
• Design, build, and scale distributed RL post-training systems, orchestrating trainer, rollout, environment, and reward workloads across thousands of GPUs.
• Build high-throughput rollout generation, integrating inference engines (vLLM, SGLang), weight synchronization, and asynchronous/off-policy schemes.
• Design RL environments for agentic, multi-step tasks — sandboxed code execution, tool use, computer use, multimodal interaction — reproducible and scalable to millions of episodes.
• Build reward infrastructure: verifiable/programmatic rewards, reward-model serving, LLM-as-judge pipelines, and defenses against reward hacking.
• Develop the evaluation, monitoring, and debugging tooling that keeps large RL runs stable.
• Advance training efficiency and stability, and turn new post-training ideas into production runs with researchers.
First 90 Days
One way the first 90 could unfold.
• Days 1–30 — Immerse & Diagnose: Learn the current RL stack and where throughput, stability, or correctness break.
• Days 30–60 — Ship & Validate: Improve a piece of the loop (rollout throughput, reward infra, or an environment) and prove it on a real run.
• Days 60–90 — Scale & Systemize: Harden the full loop across thousands of GPUs and asynchronous architectures.
What You Bring
• 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) for foundation models.
• Experience building RL environments, reward functions, verifiers, or evaluation harnesses for LLM agents, including sandboxed execution and multi-turn tool use.
• Deep familiarity with RL post-training frameworks (veRL, OpenRLHF, TRL, Ray orchestration) and rollout inference engines (vLLM, SGLang).
• Strong understanding of GPU clusters, networking, and communication libraries (NCCL, MPI) under mixed training and inference workloads.
Nice to Have
• Running RL training across 100+ GPUs, including asynchronous or disaggregated trainer/rollout architectures.
• Containerization and orchestration (Kubernetes, Ray) for large environment fleets and sandboxed workloads.
• Research contributions in RL for LLMs, or open-source contributions to RL training frameworks.
About Luma: Luma's mission is to build unified general intelligence that can generate, understand, and operate in the physical world. We believe multimodality is critical for intelligence — the next step beyond language models comes from vision. Luma is an equal opportunity employer.