Member of Technical Staff, Mechanistic Interpretability

About the role

Radical Numerics seeks a Member of Technical Staff to advance mechanistic interpretability of multimodal genome language models. You will study how these models represent, process, and reason about biological information, developing tools and experiments to open the black box and drive scientific discovery.

Key responsibilities

• Design and execute experiments to uncover features, circuits, and mechanisms driving model behavior.
• Build tooling and infrastructure for activation analysis, causal interventions, probing, and large‑scale interpretability workflows.
• Connect internal mechanisms to model capabilities, robustness, and scientific usefulness.
• Develop new interpretability techniques to improve evaluation, safety, and model design.
• Collaborate with architecture, training, systems, safety, and biology teams to translate insights into better models and outcomes.

Required profile

• Strong background in machine learning research, especially large language models and representation learning.
• Deep understanding of modern deep‑learning architectures, transformers, and large‑scale training methods.
• Proficiency in designing rigorous empirical studies and evaluating competing hypotheses.
• Excellent written and verbal communication skills for explaining technical insights.

Required skills

• Python
• PyTorch