India faces major healthcare access challenges, especially in rural areas with limited doctors and long travel times. Telemedicine bridges this gap, but current clinical history-taking at the community level is inefficient—CHWs often collect lengthy, unfocused histories before connecting with hub doctors. Our project leverages AI and Large Language Models (LLMs) to create an adaptive, efficient history-taking tool that guides CHWs to ask relevant questions based on patient responses. This system reduces consultation time, improves information quality, and helps doctors make quicker, more accurate decisions. We are developing and optimizing this tool using agent-based simulations and prompt engineering, and will evaluate its performance with both expert clinical reviewers and in a real-world pilot in rural Nashik, India. This innovation aims to enhance care quality, reduce delays, and ultimately scale across India’s public telemedicine ecosystem, including eSanjeevani.

Crema is an app for coffee lovers to share more detailed and dedicated reviews of their favorite coffee drinks and cafés. It works in a similar way to Beli, a social media community app that offers users the ability to rank, journal, and share reviews of restaurants. Crema directly focuses on the untapped market of coffee lovers and all things coffee and delivers more than just an ambiguous rating you would find on Google or Yelp.

We aim to implement The Zipper Algorithm, an elementary procedure to develop a conformal map between particular regions, in such a way to minimize error. The Zipper Algorithm repeatedly applies various transformations to the complex plane, which can result in loss of precision in points that are close together and greater distance between faraway points. This project will implement each step of the map into its elementary transformations as described in Marshall and Rohde’s paper Convergence of a Variant of The Zipper Algorithm for Conformal Mapping. A function is said to be poorly conditioned if a small perturbation in the inputs results in a large change in outputs. We aim to conduct the entire mapping within the unit disc in attempts to eliminate some of the poorly conditioned steps, and in doing so provide bounds on the error that may be introduced in such a mapping.

Opioid addiction is a huge problem in the United States that is destroying lives. Millions of Americans suffer from opioid misuse, and millions are seeking treatment through addiction-curbing medication such as naltrexone. Currently, naltrexone is delivered through an oral pill or an injected device. Patients on oral naltrexone tend to have less retention to the drug because of withdrawal symptoms, but it must be taken daily since naltrexone absorbs quickly in the body. The injected form remedies this problem through continuous release, but it is expensive, invasive, and requires periodic check-ins.​
​
We propose NaxPatch, a microneedle patch that can continuously administer naltrexone intradermally. It can be placed on the upper arm and worn discreetly. It would also be replaced once a week, which fits many schedules and is hard to forget. We believe NaxPatch will be a cheaper alternative form of naltrexone that is safe, effective, and convenient.​