Staff ML Engineer: Agent Training & Environments

Labelbox; San Francisco (CA); Hybrid; USD 250,000 - 280,000; Full time

Job summary

Labelbox is the RL data factory for advancing frontier agent capabilities. We build the data, environments, and evaluations that frontier labs use to train and judge their agents.

This role sits where training meets infrastructure. You will run the experiments and build the systems that run them: environments agents act in, verifiers that decide whether they succeeded, and the fine-tuning pipelines that turn that signal into a better model.

Qualifications

• 3 year track record of shipping systems that customers and other engineers still rely on.
• Exceptional throughput with high-quality output; your v1 is the foundation for the team.
• Strong system and API design judgment; you make hard architectural calls and defend them.
• Ship production code daily; you know where it breaks and how to fix it to move the team faster.
• Build the substrate other people’s work run on—tools, CI, harnesses, libraries.

Responsibilities

• Develop RL environments for agentic tasks, including task definitions, tool surfaces, state and reset semantics, reward design, and scalable harnesses.
• Create verifiers or graders for open-ended work, with reliable scoring pipelines.
• Build fine-tuning pipelines for SFT and RL, from data collection through training to evaluation.
• Develop eval systems that measure model and product quality across agent trajectories.
• Design and maintain training and serving infrastructure that scales with frontier labs.

Skills

Python

System design

API design

Production code

ML infrastructure

RL

Tools

GraphQL

Kubernetes

MySQL

PostgreSQL

Job description

Labelbox is the RL data factory for advancing frontier agent capabilities. We build the data, environments, and evaluations that frontier labs use to train and judge their agents.

This role sits where training meets infrastructure. You will run the experiments and build the systems that run them: environments agents act in, verifiers that decide whether they succeeded, and the fine-tuning pipelines that turn that signal into a better model.