Research Program

Advancing AI for Human Benefit

Our research aims to develop artificial intelligence systems that are effective, fair, and beneficial for society.

Research Areas

We develop models that understand and generate human language with unprecedented accuracy. Our work focuses on contextual understanding, multilingual capabilities, and handling complex discourse structures. Recent projects include large language model interpretability and cross-lingual transfer learning.

Machine Learning for Healthcare

Applying AI techniques to improve medical diagnosis, treatment planning, and patient outcomes. We work closely with clinicians at Stanford Hospital to develop systems that integrate seamlessly into clinical workflows while maintaining high accuracy and reliability.

AI Ethics and Fairness

Our research investigates bias in AI systems and develops methods to ensure fair and equitable outcomes across different demographic groups. We create evaluation frameworks, debiasing techniques, and guidelines for responsible AI deployment.

Human-AI Collaboration

Designing systems where humans and AI work together effectively. We study how to best leverage the complementary strengths of human expertise and machine capabilities, particularly in high-stakes decision-making contexts.

Education & Career

Associate Professor

Stanford University, Department of Computer Science

2016-2020

Assistant Professor

MIT, Computer Science and Artificial Intelligence Laboratory

2014-2016

Research Scientist

Google Research

Worked on machine translation and conversational AI systems.

2010-2014

Ph.D. in Computer Science

Carnegie Mellon University

Thesis: "Understanding and Generating Natural Language in Context"

2006-2010

B.S. in Computer Science

UC Berkeley

Graduated summa cum laude

Mathematical Foundations

Core Methods

Our research builds on several foundational mathematical frameworks that enable rigorous analysis of complex systems.

The fundamental optimization problem we consider is minimizing f(x) = \sum_{i=1}^{n} \ell(h_\theta(x_i), y_i) + \lambda \|\theta\|^2 where \ell is the loss function and \lambda controls regularization strength.

For deep learning models, we analyze the gradient flow dynamics governed by \frac{d\theta}{dt} = -\nabla_\theta \mathcal{L}(\theta).