Tom McGrath is Co-Founder and Chief Scientist at Goodfire, an AI interpretability research lab that decodes how neural networks think and enables precise control of AI model behavior.

Tom earned his PhD from Imperial College London and worked as a Senior Research Scientist at Google DeepMind from 2019 to 2023, where he founded their interpretability team. He also attended the University of Warwick and conducted research on LLMs, interpretability, and AlphaZero.

Tom co-authored pioneering research on the Hydra Effect and Copy Suppression in LLMs, and helped establish the field of mechanistic interpretability. His work on extracting knowledge from superintelligent models like AlphaZero has become foundational in AI safety research.

Goodfire raised a $7M Seed round in August 2024, followed by a $50M Series A in April 2025 led by Menlo Ventures, and most recently a $150M Series B in February 2026 at a $1.25B valuation led by B Capital.

Goodfire's Ember platform is the only interpretability tool capable of interpreting DeepSeek R1 and has enabled breakthrough discoveries including novel Alzheimer's biomarkers from epigenetic models. The company has reduced hallucinations by 50% in large language models through targeted interventions.