RL Infrastructure Engineer
A rare infrastructure role at the core of a frontier RL research operation. Seed-stage, well-funded, and staffed by researchers from top AI labs and PhD programs — building systems that automate task objectives and create scalable learnable curriculums across thousands of GPUs.
Frontier RL at the Earliest Stage
This is an early-stage AI research company focused on reinforcement learning in open-ended settings — building systems that automate task objectives and create scalable learnable curriculums. The team of 6–8 includes researchers from top frontier AI labs and PhD programs.
Seed-stage and well-funded, the company operates on weekly hypothesis-driven sprints. Each engineer develops a hypothesis and demos progress by end of week — a culture that rewards curiosity, speed, and rigorous thinking in equal measure.
The client organization is confidential. All conversations are handled with full discretion. Apply or reach out directly — even if you're not actively looking.
Own the Systems Layer
This is an infrastructure engineering role at the core of a frontier RL research operation. You'll build the systems layer that enables researchers and applied ML engineers to run, debug, and reproduce large-scale RL experiments — covering distributed rollouts, training orchestration, inference, evals, data pipelines, observability, and reliability.
You'll own infrastructure projects end to end, from architecture through deployment and long-term maintenance — translating messy experimental workflows into durable, scalable infrastructure alongside some of the sharpest RL minds in the field.
1 hire — urgent, target close within 1–3 months.
What You'll Build
- Infrastructure for distributed RL training and inference across thousands of GPUs
- Reliability, debuggability, and throughput improvements for large-scale RL experiments
- Interfaces for researchers and applied ML engineers to launch, inspect, compare, and reproduce experiments
- Elimination of bottlenecks in training, rollout generation, eval execution, data movement, and cluster utilization
- Engineering standards for RL infrastructure: testing, observability, versioning, and reproducibility
Technical Domain
Who You Are
- 2+ years building infrastructure for LLM or RL systems
- Experience at a high-engineering-bar organization — top AI startups, frontier labs, or Big Tech RL research teams
- Hands-on experience with GPU clusters, distributed training, model serving, or high-throughput inference systems
- Familiarity with vLLM, SGLang, and modern LLM-RL training frameworks
- Degree in CS, EECS, Mathematics, or a related field
- Very high level of curiosity and hypothesis-driven thinking
- Experience working closely with ML researchers building infrastructure for messy experimental workflows
- Evidence of strong independent technical work — open-source projects, competitions, or notable infrastructure contributions
- Familiarity with veRL, SkyRL, Slime, FSDP, DeepSpeed, or similar distributed training frameworks
Candidates who match any of the following will not be considered, regardless of other qualifications.
- No hands-on LLM or RL infrastructure experience
- No meaningful GPU or distributed systems exposure
- No evidence of strong technical ownership or independent work
Comp & Logistics
- $300,000 – $500,000 base (average offer $400K–$425K)
- Competitive equity at seed stage
- Location: San Francisco, CA (FiDi) — on-site 5 days/week
- Visa: H1B transfer, OPT, and O-1 supported
- 1 opening — urgent, target close within 1–3 months
Interview Process
AIONIA Intro Call
Background, fit, and logistics — confidential, no commitment required
Technical Interview
Deep-dive on infrastructure experience, systems design, and RL/LLM stack
On-site Loop
1–2 days on-site with researchers and engineering team in San Francisco