Mohini Anand
Projects | Publications | Contact
About
|
PhD Student, Purdue UniversityCo-Advisors: Prof. Xavier Tricoche |
I’m a PhD student in Computer Science at Purdue University, where I work on representation learning for complex, structured data, primarily in medical imaging. I’m particularly interested in problems where meaningful structure exists but is not directly observable, evaluation metrics are weak or ambiguous, and modeling assumptions matter as much as algorithmic choice.
Much of my work focuses on cardiac diffusion MRI, where hundreds of thousands of fiber trajectories must be analyzed without labels or ground truth. Over the course of my PhD, I have tested modeling assumptions against messy medical data, which has made me comfortable iterating through approaches that fail until the representation itself begins to reflect underlying structure. These settings have shaped how I approach representation learning: favoring geometric structure, domain-informed features, and interpretability over purely black-box performance.
Before starting my PhD, I earned my B.S. in Computer Science from NYU Tandon. I enjoy working across the stack, from data preprocessing and model design to interpretation and visualization, with an emphasis on producing representations that support downstream analysis and clinical insight.
Featured Projects
Unsupervised Representation Learning for Cardiac Fiber Analysis
First-of-its-kind framework combining self-supervised embeddings with anatomical context to cluster >400k cardiac fibers. Enables quantitative studies of myocardial architecture at organ-scale.
Publications/ Posters
| Year | Title | Links |
|---|---|---|
| 2025 | Deep Representation Learning for Unsupervised Clustering of Myocardial Fiber Trajectories | arXiv |
| 2025 | AI Tool to Bridge Rural Disparities (Poster) | ScholarWorks |