Labelbox
Staff ML Engineer, Agent Training & Environments
$250K–$280K
San Francisco Bay Area · Staff · Full-time · Posted Jul 29, 2026
- Experience
- 3+ years shipping systems; production SFT plus at least one RL method for agentic tasks.
About the job
Build and run reinforcement-learning environments, verifiers, evaluations, fine-tuning pipelines, and scalable infrastructure to improve and assess frontier AI agents.
What you'll do
- Design and scale agent environments, task definitions, reward systems, verifiers, and evaluation harnesses.
- Build SFT and RL fine-tuning pipelines and evaluate agent performance across training runs.
- Develop reliable training, serving, and distributed evaluation infrastructure, including orchestration and fault tolerance.
What they ask for
- Deep proficiency in Python.
- At least 3 years shipping systems relied on by customers and engineers.
- Production experience fine-tuning models for agentic tasks using SFT and at least one RL method (GRPO, PPO, DPO, or similar).
- Experience building agent environments and verifiers or graders for open-ended work.
Nice to have
- Experience with agent harnesses or coding agents as training and evaluation subjects.
- Experience in production distributed systems, ML infrastructure, or large-scale data systems; multi-tenancy and isolation for untrusted execution; or direct work with frontier labs or highly technical customers.
Pay
$250,000 — $280,000 USD (base salary)
Tools
Python, React.js, Redux, TypeScript, Node.js, GraphQL, Google Cloud Platform (GCP), Kubernetes
AI focus
Post-train and evaluate agentic AI using reinforcement learning, SFT, purpose-built environments, verifiers, and large-scale trajectory evaluation.
Model-written summary · the original posting is authoritative