Research Engineer - GPU Inference for RL Systems (Remote)

Prime Intellect AI; San Francisco (CA); Hybrid; USD 150,000 - 350,000; Full time

Benefits offered by this job

Cash compensation

Remote option

Visa sponsorship

Off-sites & learning

Mission-driven team

Job summary

Prime Intellect AI is building the open superintelligence stack, a full-stack platform for frontier-scale post-training workflows, RL and tool use. We seek engineers to optimize GPU inference, scale large models on NVL72, and push efficient RL deployment.

You’ll work on vLLM integrations, multi-node inference, FP8 precision, and collaboration with model researchers to balance quality and cost. Flexible remote or SF office and visa sponsorship are available.

Qualifications

• Experience building or optimizing GPU inference systems.
• Strong Python skills and PyTorch experience.
• Ability to profile end-to-end workloads and derive performance gains.
• Experience debugging numerical or integration issues in ML systems.

Responsibilities

• Extend prime-rl’s vLLM integration to bring new model architectures into RL training, with small-scale validation before larger runs.
• Improve rollout throughput for long-horizon agents, tuning concurrency and KV-cache reuse as context lengths grow.
• Optimize multi-node MoE inference, including expert parallelism and the tradeoffs of separating prefill from decode.
• Reduce trainer-to-inference weight-update overhead through sharding-aware transfer and communication overlap, while keeping rollout workers on the intended policy version.
• Build on FP8 and evaluate lower precision. Use trainer/sampler mismatch checks to catch numerical regressions and measure the effect on RL training.
• Work with model researchers on architecture choices that improve inference efficiency, balancing model quality against execution cost on the target hardware.

Skills

GPU inference

Python

PyTorch

Profiling

Tools

CUDA

Triton

Job description

Prime Intellect AI is building the open superintelligence stack, a full-stack platform for frontier-scale post-training workflows, RL and tool use. We seek engineers to optimize GPU inference, scale large models on NVL72, and push efficient RL deployment.

You’ll work on vLLM integrations, multi-node inference, FP8 precision, and collaboration with model researchers to balance quality and cost. Flexible remote or SF office and visa sponsorship are available.