Research

Exploring the intersection of reinforcement learning, decision making, and real-world impact.

Ongoing

Reinforcement Learning for Foundation Models

Overview Leading research efforts at MBZUAI’s Institute of Foundation Models to develop practical RL techniques for improving language model reasoning and alignment. Our work spans multiple reasoning domains through the...

Learn More

2023-2024

Safety-Critical Offline Reinforcement Learning

Overview Developing risk-sensitive methods for identifying dangerous states and treatments in healthcare settings. Focus on dead-end identification using distributional RL and conservative value estimation for improved patient safety. Motivation In...

Learn More

Ongoing

Reinforcement Learning for Foundation Models

Leading research efforts at MBZUAI's Institute of Foundation Models to develop practical RL techniques for improving language model reasoning and alignment. Our work spans multiple reasoning domains through the LLM360 initiative.

RL for LLMs Foundation Models

2023-2024

Safety-Critical Offline Reinforcement Learning

Developing risk-sensitive methods for identifying dangerous states and treatments in healthcare settings. Focus on dead-end identification using distributional RL and conservative value estimation for improved patient safety.

Healthcare Safety Offline RL

2024

Robust Autonomous Systems via Self-Play

Developed massively scalable self-play approaches for robust autonomous driving agents during my postdoc at Apple. Achieved state-of-the-art performance through aggressive randomization and large-scale parallel simulation.

Autonomous Systems Self-Play

2020-2023

Transfer Learning in Clinical Decision Support

Investigating methods for transferring learned policies across different patient populations and healthcare settings. Combining causal inference with offline RL to enable safe knowledge transfer in data-scarce domains.

Transfer Learning Causal Inference Healthcare

2022-2023

Uncertainty Quantification for Irregular Time Series

Extending evidential deep learning to continuous-time settings for handling irregularly sampled medical data. Provides well-calibrated uncertainty estimates that expand temporally for sparse observations.

Uncertainty Time Series

2017-2019

Hidden Parameter MDPs for Transfer Learning

Developed efficient transfer learning methods using hidden parameter Markov Decision Processes. Enables robust policy learning across related tasks with varying dynamics, demonstrated in robotics and control domains.

Transfer Learning MDPs