Research Interests

High-Performance Computing & ML

GPU Programming and High-Performance Computing for efficient machine learning algorithms. GPUs are major contributors to the current advancements in AI by enabling highly parallel computation. I am interested in crafting efficient algorithms both for data and computaion parallelism.

Sequence Modeling

Sequential data are ubiquitous in many shapes and forms. They may be in the form of time series or not. I am broadly interested in problems related to sequential data, namely natural language processing, video understanding, time series forecasting, robotics & autonomous vehicles, health monitoring, protein structure prediction, molecule generation or other areas.

Mathematical Modeling & Dynamical Systems

The power of mathematics in describing the world is truly mesmerizing. Dynamical systems and mathematical modeling are the backbone of many scientific and engineering disciplines. They provide a framework for understanding and predicting the behavior of complex systems over time. My interests include developing and analyzing mathematical models for various applications, specifically biological and cognitive systems. I am particularly fascinated by the interplay between theory and computation in solving real-world problems. I often find myself exploring interdisciplinary research for enhancing our understanding of complex natural systems.

Optimization

Optimization problems and mathematical programming (LPs, ILPs, MILPs, etc.), along with optimization (non-linear) for deep learning. Many of the complex algorithms and machine learning methods can inherently be looked at through the lens of optimization. I am interested in optimization under uncertainty and constraint learning as well as efficient large scale and numerical optimization. This is sometimes intertwined with my other interests such as HPC. I am inspired by the work of Professor Dimitris Bertsimas.

Computer Vision & Scene Perception

Computer vision and scene perception are the backbone of many applications in robotics, autonomous vehicles, and human-computer interaction. I am interested in the theoretical and applied aspects of computer vision, including but not limited to object detection, semantic segmentation, and scene understanding, particularly when combined with other modalities such as audio, text, and time series data. I am also interested in physics-aware video understanding and generation.

Physics-Informed ML

Levaraging the governing physical laws of phenomena for more accurate modeling and prediction. My interests span the spectrum of foundations and applications of physics-informed ML (PIML). For example ocean modeling, climate modeling, molecular dynamics, and fluid dynamics. More foundational examples include operator learning, PDE-constrained optimization, and neural architectures for PINNs. I am interested in various aspects of PIML, such as but not limited to physics-enforced optimization methods for machine learning. I am generally inspired by the work of Professor George Em Karniadakis.

Graph Learning

Graph neural networks (GNNs) are an emerging branch of machine learning that capture the relations of entities. They are particularly useful for representing complex domains such as social media and drug discovery. I am mostly interested in the theoretical foundations of GNNs, especially in combination of physics-informed learning.