Research

Brief Research Statement 

I am interested in information theory, machine learning, bioinformatics and causality. Presently, I am working on application of causal learning for sepsis patients in ICU aimed at analyzing patient outcomes such as mortality and length of stay. I am also interested in causal structure learning models with examples of real-world data to help healthcare audiences understand and apply them. 

Previously, I have worked on on causal analysis and estimation of treatment effects of various drugs from health records using machine learning techniques. Based on nationwide claim data that includes 60M patient records (Optum), we applied treatment effect methods to Alzheimer’s disease and studied the treatment effect of anti-asthmatic and anti-hypertensive drugs. 

In my Ph.D. research, I focused on the intersection of machine learning and information theory, with focus on data storage. I worked on hardware implementation of deep neural networks (DNNs) , which is an emerging area of research. When weights of neural networks are stored in hardware (e.g. non volatile memories like memristors), various types of noise will appear in the cells and degrade the performance. Additionally, there might be various kinds of random or adversarial noise in the weights or inputs of such networks. We developed codes which not only protected weights but also considered coded construction of DNNs. I also studied how to use natural redundancy (NR) in data for error correction, and how to combine it with error correcting codes to significantly improve data reliability. The complex structures of NR, however, require machine learning techniques. We study two fundamental approaches to use natural redundancy for error correction. The first approach, called Representation-Oblivious, requires no prior knowledge on how data are represented or compressed in files. It uses deep learning to detect file types accurately, and then mine Natural Redundancy for soft decoding. The second approach, called RepresentationAware, assumes that such knowledge is known and uses it for error correction

Experience

Postdoctoral Associate, Department of Surgery, School of Medicine, Duke University, NC (03/2024-)

Postdoctoral Fellow, Dept. of Biomedical Informatics, School of Medicine, Emory University, Atlanta. (12/2022 -02/2024)

Postdoctoral Research Fellow, School of Biomedical Informatics, University of Texas Health Science Center at Houston. (10/2020-09/2022)

• Causal structure learning, causal inference, heterogeneous treatment effects.

 Graduate Research/Teaching Assistant, Dept. of Computer Science and Engineering, Texas A&M University. (08/2014 - 08/2020) 

• Error Correction using Natural Redundancy.

• Machine Learning Techniques for Error Correction.

Research Associate, Dept. of Civil Engineering, Indian Institute of Technology, Bombay,  08/2011-07/2013

• Composite Signal Control Strategy (CoSiCoSt-2G).

• Simtram, a traffic simulator for Indian road conditions.

 Research Intern, Department of Computer Science and Engineering, Indian Institute of Technology, Guwahati (05/2010- 07/2010)

• Implementation of Bloom Filter based IP Lookup algorithm in Click Modular Router.