Enric Boix-Adsera

Assistant Professor of Statistics and Data Science
Wharton School at University of Pennsylvania

eboix@wharton.upenn.edu

CV (pdf)

People / Teaching / Papers

My research focuses on building a mathematical science of AI. My goal is to understand fundamental theoretical principles in order to enable more efficient and trustworthy AI systems. Recently, some topics that I have been thinking about are:

  1. AI safety: How can we ensure AI systems are trustworthy as their capabilities improve? How can we monitor and control their behavior?
  2. AI reasoning: How can we make networks that generalize well to outputs outside of their training distribution? Why should we expect this to be possible at all?
  3. AI representations: What's going on inside a neural network? What are the fundamental principles driving how it learns internal representations that help it generalize? What makes an internal representation "useful"?
These questions are richly interconnected, and progress on one can inform the others. Beyond these themes, I am always happy to chat about new ideas and learn about new paradigms!

People

I am very fortunate to work with many amazing people, including the students in my group:

The group is growing! I am looking to hire creative students who have strong mathematical and empirical AI research skills. If you are interested in working with us, please see this FAQ!

Teaching

I designed a new course, which I am teaching for the first time in Spring 2026:
STAT 4850/5850 Foundations of AI: Deep Learning with Applications.

Papers

* denotes equally-contributing first authors and (αβ) denotes alphabetical order

AI Safety and Interpretability
AI Reasoning
How does AI learn internal representations? (Deep Learning Theory)
Optimal transport
Provable machine learning algorithms
Other

Website design adapted from Tengyu Ma's