Detailed and historical land records are rarely available easily, especially in India. This restricts many at-scale solutions that depend on quality data being available.

Under the DILRMP project, the government has been working on digitizing Indian land records. I've been able to scrape and process this data for several states and am working towards releasing an open-source dataset.

Under the Ayushman Bharat program, the government has been rapidly digitizing healthcare records across the country. A vast set of stakeholders have different levels of access to this data.

Much of this data is "anonymized" by removing identifiers like names and addresses. I demonstrate in an upcoming manuscript that this data can be easily de-anonymized, posing grave privacy concerns.

In football, the probability of scoring goals is measured by a metric called expected goals (xG). One of the leading providers of xG, Statsbomb, makes a small amount of data open-source.

Using this data, I was able to recreate the industry-leading xG model and add some bells and whistles (interpretability, form-adjustment, etc.)

A small web game based on the Chrome Dino that I made for the deluxe release of Bagi Munda's Player No. 1 over the summer of 2025. Funnily enough, the guy who commissioned it (Jai Kher of 5.5 Records) never paid me, but it was a good learning curve.

One of the first projects I made when I was still learning to code. Mostly interesting because I coded the maze generation and solving (A*) algorithms from scratch. First time I realized how tough the sprite business is.