Research, Coding Agents

Thinking Machines Lab Inc.; San Francisco (CA); On-site; USD 350,000 - 475,000; Full time

Benefits offered by this job

Health, dental, and vision benefits

Unlimited PTO

Parental leave

Relocation support

Job summary

Thinking Machines Lab Inc. is seeking a researcher to lead RL training for agentic coding capabilities. You will design and run training jobs, iterate on recipes and data, and build scalable tooling for data generation and evaluation.

You will also develop sandboxed coding environments, curate synthetic data, and implement evaluations to drive real-world improvements. Collaboration with infra, evals, and post-training teams will ship results into model releases.

Qualifications

• Strong engineering skills and ability to contribute code and debug.
• Proficiency in Python and familiarity with at least one deep learning framework (PyTorch, TensorFlow, or JAX).
• Bachelor’s degree or equivalent in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding.
• Clarity in communication, an ability to explain complex technical concepts in writing.
• Experience building synthetic data pipelines and systems adopted by teammates.
• Experience owning end-to-end cycles to close gaps in model usability via custom evaluations and training data.
• Experience making large-scale agentic RL infrastructure reliable.
• PhD in related field or equivalent industry research experience.

Responsibilities

• Design and run RL training jobs targeting agentic coding capabilities, iterating on recipes and data.
• Build and improve the sandboxed coding environments and reward signals that models are trained and evaluated against.
• Generate and curate high-quality synthetic coding data, and build scalable, general-purpose data pipelines.
• Design evals that measure real-world coding usefulness, and train models against them to deliver concrete improvements in day-to-day usability.
• Debug and analyze large RL runs to catch confounders, reward hacking, and other RL failure modes.
• Collaborate closely with infra, evals, and other post-training teams on shared data, joint training runs, and usability improvements — and ship the results into model releases.

Skills

Python

DL Frameworks

Distributed Training Debugging

Code Debugging

Strong Engineering Skills

Communication

Education

Bachelor’s degree or equivalent in CS/ML/Physics/Math

PhD in CS/ML/Physics/Math or equivalent industry research

Job description

About Thinking Machines

The mission of Thinking Machines is to build AI that extends human will and judgment. We are training frontier models with Inkling, developing Tinker to let people make models their own, and crafting interfaces that broaden human-AI communication. We believe the future worth building is human, and we're hiring people who want to build it.

About the Role

• The Coding Agents team makes our models world-class at agentic coding — writing, debugging, and reasoning about code across long-horizon, multi-turn tasks.

• You'll join a small, high-leverage team responsible for the recipes, data, and infrastructure behind coding capability gains in every model release.

• The team owns the full coding post-training stack: synthetic and human data generation, RL environments and sandboxes, reward and grading design, and large-scale training runs.

• This is a research role with real ownership — you'll shape technical direction, not just execute against a spec.

What You’ll Do

• Design and run RL training jobs targeting agentic coding capabilities, iterating on recipes and data.

• Build and improve the sandboxed coding environments and reward signals that models are trained and evaluated against.

• Generate and curate high-quality synthetic coding data, and build scalable, general-purpose data pipelines.

• Design evals that measure real-world coding usefulness, and train models against them to deliver concrete improvements in day-to-day usability.

• Debug and analyze large RL runs to catch confounders, reward hacking, and other RL failure modes.

• Collaborate closely with infra, evals, and other post-training teams on shared data, joint training runs, and usability improvements — and ship the results into model releases.

Skills and Qualifications

Minimum qualifications:

• Strong engineering skills, ability to contribute code and debug in complex codebases.

• Proficiency in Python and familiarity with at least one deep learning framework (e.g., PyTorch, TensorFlow, or JAX). Comfortable with debugging distributed training and writing code that scales.

• Bachelor’s degree or equivalent experience in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding.

• Clarity in communication, an ability to explain complex technical concepts in writing.

• Experience building synthetic data pipelines and systems that were adopted by others on your team and remain in use today.

• Experience owning the end-to-end cycle of identifying gaps in model usability and closing them through custom evaluations and training data.

• Experience making large-scale agentic RL infrastructure reliable given the long tail of failures that surface at scale.

• Experience improving the coding capabilities of a frontier model.

• PhD in Computer Science, Machine Learning, Physics, Mathematics, or a related discipline with strong theoretical and empirical grounding; or, equivalent industry research experience.

Logistics

• Location: This role is based in San Francisco, California.

• Compensation: Depending on background, skills and experience, the expected annual salary range for this position is $350,000 - $475,000 USD.

• Visa sponsorship: We sponsor visas. While we can't guarantee success for every candidate or role, if you're the right fit, we're committed to working through the visa process together.

• Benefits: Thinking Machines offers generous health, dental, and vision benefits, unlimited PTO, paid parental leave, and relocation support as needed.