RL Environments Architect: Scale & Automate Task Creation

Bespoke Labs Inc.; Mountain View, Northern (CA, KY); Hybrid; USD 250,000 - 300,000; Full time

Benefits offered by this job

Health, dental, and vision coverage

401(k)

Daily onsite lunch provided

Visa sponsorship and relocation relief

Direct impact on industry training and

Job summary

Bespoke Labs Inc. seeks an engineer to build the machinery that turns environment ideas into hundreds or thousands of validated agentic coding tasks.

You will design pipelines that mass-produce environments, craft complex coding worlds around real codebases, and push throughput with minimal manual work per task. We measure you on volume and quality of environments shipped, not papers, and you will directly influence frontier coding agents and their validation, with strong emphasis on scalable

Qualifications

• Experience building agentic coding tasks or environments and quantifying volume and cost.
• Experience scaling output through automation rather than more people.
• Strong software engineering fundamentals and familiarity with production-grade code in multiple languages.

Responsibilities

• Build environment-generation pipelines to produce RL environments at scale with templating, grading, verification, and QA.
• Create high-fidelity coding worlds around real codebases with proper tooling and dependencies.
• Scale task creation to thousands with automation handling the heavy lifting.
• Build internal tools and infrastructure to raise throughput and remove bottlenecks.
• Own the full task lifecycle from prompt to evaluation, ensuring rigor and fairness.
• Defend quality at scale by catching reward hacking and grader loopholes.

Job description

Bespoke Labs Inc. seeks an engineer to build the machinery that turns environment ideas into hundreds or thousands of validated agentic coding tasks.

You will design pipelines that mass-produce environments, craft complex coding worlds around real codebases, and push throughput with minimal manual work per task. We measure you on volume and quality of environments shipped, not papers, and you will directly influence frontier coding agents and their validation, with strong emphasis on scalable