Prime Intellect

Research Engineer building the systems infrastructure that enables large-scale reinforcement learning and distributed model training. The role focuses on performance optimization across GPUs, kernels, networking, scheduling, and runtime systems for frontier-scale AI.

Build the data foundation for Prime Intellect’s next generation of open AI models by sourcing, curating, and evaluating large-scale training corpora. You’ll combine scalable data engineering with research-driven experiments across pretraining, reasoning, coding, and long-context data.

Build the tasks, environments, data pipelines, and evaluations that train and measure frontier AI models through reinforcement learning. This role combines hands-on Python engineering with investigation of model failures, reward hacking, and agent behavior.

Research Engineer - Pretraining will build Prime Intellect’s next open language models, taking architecture experiments through full-scale training runs of up to 1–3T parameters. The role combines frontier model research with hands-on training systems engineering, optimization, and reliability work.

Build high-performance inference systems that power Prime Intellect’s frontier RL research and scale toward 1–3T parameter models. You will optimize rollout throughput, distributed MoE inference, precision, and trainer-to-inference updates on advanced GPU clusters.

Build and operate high-throughput storage infrastructure powering frontier AI training, inference, and research workloads. This role owns performance, reliability, durability, automation, and cost as GPU clusters scale.

Prime Intellect is seeking a Founding Solutions Architect to turn customer AI workloads into benchmarked, production-ready architectures on its frontier AI infrastructure. The role owns technical discovery, proofs of concept, competitive benchmarking, and the creation of scalable solutions playbooks.

Build Prime Agent, an open-source coding and research agent, as a generalist software engineer spanning intuitive product interfaces and underlying agent systems. You’ll own features end-to-end in person from Prime Intellect’s San Francisco office.