Hello, I’m Gerard Boxó.
I am a researcher based in Barcelona, specializing in the intersection of Machine Learning and Biology. My primary work focuses on developing and applying ML methods to protein engineering, leveraging techniques ranging from Reinforcement Learning on protein language models to Mechanistic Interpretability.
Concurrently, I conduct research on critical problems in AI Alignment and Control. I have found deep methodological connections between AI Safety and AIxBio—specifically in interpretability, rigorous evaluations, and applied RL.
This blog serves as a repository for my research journey, where I publish experiments, analyses, and insights.
Research Interests
Mechanistic Interpretability
Reverse-engineering neural networks to understand their internal cognition.
AI x Biology
Applying interpretable ML to protein engineering and drug discovery.
AI Safety
Ensuring advanced AI systems remain aligned and controllable.
Current Focus
Current Research
I am currently focused on evaluating the capabilities of frontier models for autonomous protein design.Education
MSc in Bioinformatics
Universitat Autònoma de Barcelona (UAB)
Focus: Computational Biology, Machine Learning
BSc in Statistics
Universitat Politècnica de Catalunya (UPC), Barcelona
Focus: Statistics, Data Science
Curriculum Vitae
For a detailed look at my experience, skills, and publications.