Staff ML Engineer: Frontier Agent Training & Environments

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

Benefits offered by this job

Hybrid work model (3 days in office)

Career advancement opportunities

Job summary

Labelbox, based in San Francisco, is the RL data factory for advancing frontier agent capabilities. We build environments, verifiers, and pipelines that train and judge agents, while providing scalable training and serving infrastructure.

This role demands strong system and API design judgment and a track record of shipping reliable production code. We seek engineers who move fast in startup contexts, own critical subsystems, and collaborate across frontend, backend, and ML infrastructure.

Qualifications

• 3+ year track record of shipping systems used by customers and engineers.
• High throughput production coding with low latency and reliability.
• Strong system and API design judgment; able to make hard architecture calls.
• Production code is shipped with agents daily and made reliable for the team.
• Build foundational tooling, CI, and harnesses for the team.
• Able to move fast in ambiguous startup environments.
• Deep proficiency in Python and comfort across the rest of the stack.

Responsibilities

• Design RL environments and task definitions for agentic tasks.
• Develop verifiers and graders with rubric pipelines and pass@k scoring.
• Build fine-tuning pipelines from data collection through evaluation.
• Run eval systems measuring model and product quality.
• Develop scalable training and serving infrastructure for high-throughput workloads.

Skills

Python

API design

System design

Production code

Tools

React.js

TypeScript

Node.js

Python

GraphQL

Kubernetes

MySQL

PostgreSQL

Job description

Labelbox, based in San Francisco, is the RL data factory for advancing frontier agent capabilities. We build environments, verifiers, and pipelines that train and judge agents, while providing scalable training and serving infrastructure.

This role demands strong system and API design judgment and a track record of shipping reliable production code. We seek engineers who move fast in startup contexts, own critical subsystems, and collaborate across frontend, backend, and ML infrastructure.