Jobs at Prime Intellect

Embed with customers to design and run RL/post-training workflows: build custom environments, evaluation harnesses, and training runs on the Lab stack; translate field insights into platform improvements and reference implementations.

This role involves designing AI agents, building robust infrastructure, and translating customer insights into technical requirements while working with reinforcement learning and applied data.

Develop and implement advanced reinforcement learning methods and distributed systems for AI agents. Collaborate with customers to understand needs and create tailored solutions.

Build and optimize GPU inference systems powering reinforcement learning research at frontier scale. Responsibilities include extending vLLM integrations, improving long-horizon rollout throughput, optimizing multi-node MoE inference, reducing trainer-to-inference update overhead, evaluating low-precision inference, debugging numerical issues, and collaborating with researchers on inference-efficient model architectures.

Build scalable pipelines that collect, clean, deduplicate, filter, and evaluate web, code, and technical training data. Develop data-quality and contamination checks, test data mixtures and synthetic data, support long-context, reasoning, and coding capabilities, and run controlled experiments to measure model impact. Ensure large-scale processing is efficient, reproducible, recoverable, and versioned while collaborating closely with AI researchers.

Build and optimize infrastructure for large-scale reinforcement learning and distributed model training. Responsibilities include improving compute, memory, networking, scheduling, kernels, communication paths, runtimes, and parallel training systems. The role contributes to RL architecture, asynchronous rollouts, post-training systems, open-source infrastructure, and frontier-scale AI training while collaborating with researchers and infrastructure engineers.