Chinmay Joshi

I am interested in machine learning systems that learn efficiently from imperfect, limited, or sensitive data. My interests sit around data-centric deep learning, trustworthy ML, and computer vision, with a focus on how data, optimization, and model behavior interact.

Recently, I have been studying dynamic data pruning, where the goal is to decide which examples a model should learn from, when, and with what influence. I am especially interested in using such training dynamics to reduce computation, limit harmful memorization, and make deep learning models more reliable in practice.

When I'm not distracted by these endless academic pursuits, I'm busy pondering philosophy or trying to predict the next Wimbledon champion. Sometimes, you might find me in the mountains with my camera or deep in the forest, lost in the world of birds.