Goodfire is an AI interpretability research company focused on understanding and designing AI systems that people can trust. The business problem is to advance humanity's understanding of AI to build safe and powerful AI systems, requiring deep research breakthroughs to make this possible.
Requirements
- 5+ years of experience in ML infra, product engineering, or systems programming.
- Expertise in Python, PyTorch or Jax, and distributed systems.
- Prior work on model internals, explainability, or interpretability.
- Open-source ML infra contributions.
Responsibilities
- Ship research into production: take an interpretability method from paper/prototype to a reliable service with clear APIs, monitoring, and docs.
- Own backend systems: design and implement services, pipelines, and storage for large‑scale interpretability workflows (Python, PyTorch, distributed infra).
- Build for real users: understand user workflows and turn experimental techniques into tools people use routinely
- Leave things better: create reusable components, benchmarks, and runbooks others rely on.
Other
- Comfort working across research and engineering boundaries.
- Comfort moving from idea to prototype and iterating fast on user feedback to get to production
- Clear written and verbal communication; can explain decisions and trade‑offs to technical and non‑technical partners.
- Enthusiasm for making AI systems safer and more interpretable.
- Startup or lab experience in fast-moving teams.
