Hi, I'm Nachiketa Patil

MS by Research Student at CMInDS, IIT Bombay

I'm an MS by Research student at CMInDS, IIT Bombay, advised by Prof. Arjun Bhagoji. I work on robust ML, adversarial training, and data-centric AI.

My thesis, Mycroft, is an efficient external data selection protocol for collaborative ML settings where multiple data owners can't fully share their data (under review, TMLR). I've also worked on ChronoSSM, which jointly trains event and timing prediction in autoregressive SSMs rather than the standard two-stage decoupled approach (KDD’27), and RepAlign, which improves adversarial training by aligning intermediate feature representations to simultaneously boost clean and robust accuracy (ICLR’27).

During my summer research internship at Motilal Oswal Financial Services Limited, I worked on CandleDiff, a diffusion model that generates structurally valid OHLC candlestick bars conditioned on firm fundamentals and market context, and contributed to a large-scale benchmark study on time-series foundation models across Indian and US equity markets — both under review at ICAIF’26.

Along with my research, I serve as the Department System Administrator at CMInDS, managing the HPC cluster, mail servers, and departmental services. I'm also a Teaching Assistant for DS 203: Programming for Data Science (Autumn ’26).