SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 150,000 - 350,000 / annual
Prime Intellect is building the open superintelligence stack—infrastructure that frontier AI labs build internally, now available to every ambitious AI team. The platform, Lab, unifies compute, environments, evaluations, secure sandboxes, high-performance training, and deployment for post-training at frontier scale, from SFT and RL to tool use, agent workflows, and continuously improving production models.
As a Research Engineer - Inference, you will own the inference systems powering Prime Intellect's RL research, scaling toward 1–3T parameter models on NVL72 clusters. Your work will directly improve rollout throughput and make frequent model updates practical for long-horizon autonomous research tasks.
Key responsibilities include:
- Extending prime-rl's vLLM integration to bring new model architectures into RL training, with small-scale validation before larger runs
- Improving rollout throughput for long-horizon agents by tuning concurrency and KV-cache reuse as context lengths grow
- Optimizing multi-node MoE inference, including expert parallelism and prefill/decode separation tradeoffs
- Reducing trainer-to-inference weight-update overhead through sharding-aware transfer and communication overlap, while maintaining rollout workers on the intended policy version
- Building on FP8 and evaluating lower precision approaches, using trainer/sampler mismatch checks to catch numerical regressions and measure RL training effects
- Collaborating with model researchers on architecture choices that improve inference efficiency, balancing model quality against execution cost on target hardware
Prime Intellect has raised $150M from Founders Fund, Radical Ventures, NVIDIA, and exceptional operators including Andrej Karpathy, Dwarkesh Patel, and leaders from Ramp, Perplexity, Harvey, Cognition, OpenAI, Databricks, and others. The company is looking for people who want to build at the intersection of frontier research, real infrastructure, and go-to-market for a category that does not fully exist yet.
Benefits include flexible work arrangements (remote or in-person at San Francisco offices), visa sponsorship and relocation assistance, quarterly team off-sites, hackathons, and conference/learning opportunities.
REQUIREMENTS:
- Experience building or optimizing GPU inference systems, with understanding of inference-engine internals
- Strong Python skills and experience with PyTorch or comparable ML framework
- Ability to profile end-to-end workloads and turn measurements into verified performance improvements
- Experience debugging numerical or integration issues in ML systems
Nice to have:
- Contributions to vLLM or another inference engine
- CUDA or Triton kernel work, particularly for low-precision inference
