About

I am an applied research scientist with a strong interdisciplinary background spanning computer science, statistics, and mathematics. I currently work on reinforcement learning (RL), LLMs, and diffusion generative models at the Harvard Computer Science Department and the Harvard Data Science Initiative.

I am passionate about building AI systems that can be deployed in real-world settings. I welcome discussions about collaborations and career opportunities in these areas.

Recent projects I have worked on include using RL to enhance LLM agent performance and interpretability, developing new transfer learning techniques for diffusion LLMs, improving the robustness of online RLHF with LLM judges, and training foundation models for counterfactual inference. My work has been published in top-tier computer science conferences and scientific journals.

Background

My journey in AI began as a data scientist in industry prior to my Ph.D., where I developed a passion for building robust, scalable, deployable systems. One of my favorite projects was developing a sequential decision-making system for automated portfolio optimization, which was deployed in production at an investment bank in Mexico. This system integrated machine learning, optimization, and expert views via Bayesian inference.

This practical foundation led me to pursue a Ph.D. at UT Austin to work on RL, machine learning, and robotics, building on my mathematics background (BSc from ITAM, master’s from Cambridge). Key projects included developing an efficient real-time deep-learning vision system for autonomous soccer robots that competed in the RoboCup competition, and creating a new algorithm for goal-conditioned RL. I also developed statistical models used by UT Austin to issue COVID-19 response recommendations to government officials in Austin, Texas. I further expanded my research experience through internships at Meta AI (FAIR) and Intel AI, working on RL and computer vision, alongside numerous collaborations with experts across diverse fields.

The thread connecting all my work is a commitment to building AI systems that facilitate optimal decision-making in real-world applications. Ultimately, I aim to develop data-driven systems that can operate autonomously and smartly in both digital and physical worlds.

2022 Robocup competition, Bangkok, Thailand