Member of the Technical Staff, Interpretability at Output

📋 Description

• Probe and reverse-engineer the model's representations across molecular scales.
• Design experiments to map model capabilities and biological function.
• Build methods to extract the model's biological understanding as usable outputs.
• Create tools linking internals to biological concepts for interpretation.
• Collaborate with pretraining and generation teams to feed findings back into development.
• Own end-to-end pipeline from probing to production-grade interpretability tools.

🎯 Requirements

• PhD in CS/ML/physics/math with 2+ years postdoc/industry research in interpretability; or BA/MA with 5+ years.
• Strong publication record at top venues on mechanistic interpretability or probing.
• Hands-on experience analyzing internal representations of large neural networks and designing experiments.
• Proficient in Python and PyTorch; experience with large models on GPU.
• Ability to translate interpretability research into usable tools and production-grade code.
• Production-quality, well-tested code; comfortable with version control and code reviews.
• Bonus: chemistry/biology background; ML on scientific data; visualization tools.

🎁 Benefits

• We encourage new ideas, creativity and contrarian thinking.
• Healthy feedback-focused environment with constructive feedback.
• You own your day-to-day management; aim for milestones.
• Competitive salary and equity in a growing startup.
• Excellent medical, dental, and vision coverage.