research

Poster and Presentation Sessions

Research Experiences

Moncrief Summer Intern, Computational Visualization Center – Oden Institute for Computational Engineering and Sciences

PI: Dr. Chandrajit Bajaj
June 2025 - Present

As a Moncrief Summer Intern, I worked on the development of computational frameworks for early and accessible detection of Parkinson’s disease. Our project focused on designing progressive AI approaches for multi-modality differential diagnosis, combining gait, imaging, clinical, and blood-based biomarkers into a computational pipeline.

Key aspects of the work include:

  • Investigating the role of CSFSAA as a diagnostic marker for PD, while addressing its limitations as an invasive and less accessible test.
  • Exploring a non-invasive alternative by integrating wearable-derived gait features, clinical assessments, and blood-based biomarkers.
  • Training a gradient boosting classifier on baseline data from PD patients in the PPMI dataset to evaluate both individual and combined feature group performance.
  • Demonstrating that the combined multimodal model achieved high accuracy (AUC: 0.931 ± 0.065), with cognitive tests, sleep scores, and gait/arm swing features contributing most.
  • Applying visualization and co-clustering methods to identify patient subgroups with shared patterns and CSFSAA outcomes.
  • Supporting the hypothesis that multimodal, remotely accessible data can predict CSFSAA positivity, enabling earlier diagnosis without invasive testing.

This ongoing research project aims to bridge computational science with clinical applications, contributing to the long-term goal of precision medicine for neurodegenerative disorders.

Skills: Using Texas Advanced Computing Center (TACC) supercomputers (specifically Vista and Lonestar6), machine learning, project management.

Undergraduate Researcher, Precision Medical Instrumentation Design Lab - Penn State University

PI: Dr. Jason Moore
January 2025 - Present

At the Precision Medical Instrumentation Design Lab, we engineer medical simulators for training high-risk but critical procedures such as colonoscopies, central line insertions, and laparoscopies. My work specifically focuses on enhancing our simulation platforms with computer vision and machine learning to improve the feedback provided to the user and increase accessibility by reducing the need for costly sensors and tools. Our ultimate goal is to give clinicians and care-givers the confidence to act decisively when every second counts during critical procedures.

Key aspects of the work include:

  • Designing a camera-based autocalibration system for accurate instrument tracking within the simulator.
  • Implementing computer vision algorithms to monitor tool motion and identify deviations from correct procedural technique.
  • Developing real-time feedback mechanisms that guide trainees through each step of the procedure.
  • Integrate vision-based tracking and feedback into a simulation platform to improve training quality and enhance patient safety.

This work aims to deliver a next-generation medical training tool that enhances skill acquisition, increases accessibility, and ultimately contributes to better patient outcomes through evidence-based simulation technology.

Undergraduate Researcher, DIMACS (Center for Discrete Mathematics and Theoretical Computer Science) - Rutgers University

PI: Dr. Pierre Bellec
May 2024 - July 2024

At Rutgers DIMACS, I conducted theoretical and computational research on statistical mechanics models and stochastic processes, focusing on the Coupling from the Past (Propp & Wilson, 1996) algorithm as a rigorous framework for perfect sampling of Markov chains. This work provided insights into the simulation of phase transitions and equilibrium states in complex systems.

Research contributions included:

  • Applying Coupling from the Past (CFTP) to achieve perfect simulation of the Ising model in statistical mechanics.
  • Optimizing the CFTP method to overcome convergence issues common in ordinary MCMC approaches.
  • Demonstrating how perfect sampling provides more accurate insights into ferromagnetic phase transition phenomena.
  • Laying groundwork for extending CFTP-based simulation to more sophisticated models in statistical mechanics with real-world relevance.

This work highlights the potential of CFTP not only to advance statistical mechanics through exact simulations of phase transitions, but also to expand into other fields where perfect sampling can address complex modeling challenges.