Staff ML Engineer, Agent Training & Environments
Core
Building RL data factories, environments, verifiers, and fine-tuning pipelines to train and evaluate frontier AI agents.
Role type
Staff ML Engineer (Agent Training & Environments)
Builds
RL environments, verifiers, graders, fine-tuning pipelines (SFT/RL), and evaluation systems for agent trajectories.
Domain
Artificial Intelligence / Reinforcement Learning / Agent Development
Deliverable
production ML models
Required skills
Python, system and API design, production code shipping with coding agents, RL post-training (SFT, GRPO, PPO, DPO), environment design, verifier/grader design, compute-economics reasoning, distributed systems, ML infrastructure.
Preferred skills
Agent harnesses and coding agents, multi-tenancy and isolation (sandboxing, egress control), production distributed systems, ML infrastructure, data systems at scale, frontier lab customer experience.
Technologies
Python, Node.js, TypeScript, React.js, Redux, GraphQL, Google Cloud Platform (GCP), Kubernetes, MySQL, Spanner, PostgreSQL, Kafka, PubSub, Java, Kotlin.
Responsibilities
Design and build RL environments for agentic tasks including task definitions, tool surfaces, state/reset semantics, and reward design. Develop verifiers and graders using programmatic checks, LLM judges, and rubric pipelines to determine agent success at scale. Construct fine-tuning pipelines converting evaluation signals into model improvements via SFT and RL. Operate eval systems running millions of agent trajectories to measure quality. Build training and serving infrastructure supporting multi-launcher orchestration, fault tolerance, and cost accounting.
Seniority
Staff, technical direction & hands-on execution
