Sid Gupta

I hope to continue research in machine learning, interpretability, and computer vision for healthcare. To date, my research and projects encompass:

1. Combining multiple modalities of data to emulate clinical predictions

2. Making black-box ML models more interpretable to understand

3. Applying ML algorithms in real-world settings to improve health in an accessible way

Machine Learning to Automate the Visual Interpretation of Chemical Agglutination Tests

S. Gupta , A. Sklavounos , J. Dahmer , A. Yong , M. Abdullah , G. Camacho , K. Morton , M. Shiu , J. Labrecque , T. Veres , A. Wheeler , A. Mariakakis
In peer-review submission

Contour-guided Image Completion with Perceptual Grouping

M. Rezanejad* , S. Gupta* , C. Gummalaru , R. Marten , J. Wilder , M. Gruninger , D.Walther
Published to BMVC 2021

Digital Microfluidic Hemagglutination Assays for Blood Typing, Donor Compatibility Testing, and Hematocrit Analysis

A. Sklavounos , J. Lamanna , D. Modi , S. Gupta , J. Callum , A. Mariakakis , A. Wheeler
Published to Clinical Chemistry 2021

Aphrodite Data Science Lead

Aphrodite is a university-wide matchmaking service that to date has matched over 40,000 students. I work with a talented group of volunteers to:

1. Apply ML on a student dataset to gain insights on relationship science and improve matching

2. Coordinate matching of students using a robust matching algorithm

3. Develop insightful data reports that build community and highlight similarities among groups of people

Team: https://aphrodite.global/about

Explaining eye-tracking CNNs

We apply DeepDream and SmoothGrad on CNNs for eye-tracking, and explain some of their black-box behavior

Team: Sid Gupta* , David Canagasabey*, Melissa Halim Lee*

Visualizing Fairness in ML

We visualize latent embeddings and weights of different ML models on an unfair dataset, comparing fair generative models, debiasing models, and unfair models

Team: Sid Gupta* , Tina Li*, Melissa Hu*

CoughMotion

We propose a CNN model that intakes audio and IMU data in two separate branches, and outputs a cough detection prediction

Team: Sid Gupta , Alex Mariakakis (Supervisor)