Research Scientist / Engineer – Reinforcement Learning Infrastructure

Lumaai is seeking a full-time Research Scientist or Engineer in London to construct the distributed systems that enable reinforcement learning at a frontier scale. The daily work involves orchestrating training, rollout generation, environment execution, and reward computation across thousands of GPUs to ensure these processes run concurrently with speed, stability, and correctness. You will design high-throughput rollout pipelines by integrating inference engines like vLLM and SGLang, while also managing weight synchronization and asynchronous schemes. A significant portion of the role focuses on creating scalable, reproducible environments for agentic tasks, including sandboxed code execution, tool use, and multimodal interaction. Additionally, you will build reward infrastructure that combines programmatic verifiers, reward-model serving, and LLM-as-judge pipelines, while implementing defenses against reward hacking. The position requires developing robust evaluation, monitoring, and debugging tools to maintain the stability of large-scale runs and to translate new post-training ideas into production. Candidates must have direct experience operating reinforcement learning at real scale, specifically with post-trained large language models, and should be familiar with debugging asynchronous rollout pipelines. The ideal person has previously built environments and verifiers and understands the complexities of keeping a full-loop system fast and correct. This role suits individuals who have lived through the challenges of scaling RL systems and are ready to deepen their expertise in this complex domain.