Tested, designed, and developed backend APIs of WebRTC enabled multi‑party video conferencing web application and delivered the project 15 days ahead of schedule by efficiently designing the flow of the project.
Optimized image processing algorithm of pill detection by creating customized MASKRCNN algorithm, increasing accuracy by 15%; trained classification algorithm with the help of triplet loss to learn the image embedding of pill, reducing the hassle of collecting data of new pills.
Devised a pipeline of the project to incorporate it into the product of the company. Implemented Restful APIs in Django that enabled our quality Analyst team to increase reporting speed by 46%.
Built a web app to onboard data from the company’s product using Flask, Postgres, and AWS S3, enabling interactive charts in real-time.
Mentored 2 interns to optimize the pill detection algorithm and to include the multiprocessing pipeline, increasing overall speed by 75%.