Research Engineer (Training & RL)

About the Role

As a Research Engineer at humans&, you'll work on the algorithms and methods that train our models. You'll run large-scale experiments, implement and improve training techniques, and push the boundaries of what's possible with reinforcement learning for language models.

This role sits at the intersection of research and engineering. You'll collaborate closely with researchers to turn ideas into working systems, iterate quickly on experiments, and ship improvements that make our models more capable. The best candidates are comfortable diving into complex ML codebases, have strong intuitions about what makes training runs succeed or fail, and are energized by working at the frontier.

We're hiring multiple engineers for this team.

What You'll Do

• Run and analyze large-scale training experiments

• Implement and improve training algorithms and methodologies

• Debug training runs—diagnose instabilities, find the source of regressions, and fix them

• Collaborate with researchers to translate research ideas into production training code

• Build tooling for experiment tracking, analysis, and reproducibility

• Contribute to the feedback loop between research insights and model improvements

What We're Looking For

• Strong software engineering skills and proficiency in Python

• Experience with deep learning frameworks (PyTorch, JAX) and large-scale model training

• Ability to iterate quickly and comfort working in complex ML codebases

• Good intuitions about training dynamics and debugging ML systems

• A bias for action and strong ownership over outcomes

Highly valued:

• Experience training large language models

• Background in reinforcement learning research or engineering

• Experience running training at scale across many GPUs

• Track record of shipping improvements to production training pipelines

• Publications or open-source contributions in ML