Mansi Khemka

À propos

I am a Machine Learning Researcher. My area of interest is in the application of ML to solve real world problems. Back in India, I had the chance to work with skilled professors who helped me nurture my research skills. One of my past projects aimed at building a light-weight ML model that could detect and classify vehicles in Indian traffic which is way more challenging than western countries traffic. This project is now deployed by AI companies in India at several traffic intersections to gather data that helps the government formulate traffic related policies like pollution control. My second paper involved modifying the Hoeffding bound for detecting changes in data distributions to work on multivariate, non i.i.d. data and using it to formalize context changes detection in extremely long videos as a concept drift detection problem. This could be an accurate yet light-weight pre-processing step to chunking long videos into smaller chunks, easily digestible by task specific, downstream ML algorithms.

It feels too early to streamline my focus to a singular area. I constantly explore new and challenging domains. My latest paper involved the study of AI on developer productivity through an internally conducted survey at Microsoft. I continue to dabble in other research projects involving Explainable AI (XAI) and Natural Language Process (NLP) and stay ahead of cutting-edge research through participation as peer reviewer in top journals and conferences.

Ultimately, I would like to harness my ML skills in conjunction with my Software Development skills to build innovative solutions that impacts millions of customers.