Hi! I’m Aditi, a second-year Computer Science Ph.D. student at McGill University and MILA – Quebec AI Institute, advised by Prof. Siva Reddy and Prof. Golnoosh Farnadi. My research focuses on bridging pretraining and post-training in language and multimodal models, particularly Mixture-of-Experts (MoEs).
Before starting my Ph.D., I spent two years at Microsoft in the Turing Team, working with Prof. Monojit Choudhury. There, I worked on ethical reasoning and pluralistic human-values alignment in LLMs, contributed to building Bing Copilot Chat, and trained state-of-the-art Visual Document Understanding models.
I earned my bachelor’s degree in Electrical Engineering at the Indian Institute of Technology Delhi. During my undergraduate studies, I interned at Microsoft, Amazon, and Siemens, working on various aspects of building and deploying real-world machine learning systems, and received the Best Bachelor’s Thesis Award for my work on scaling Graph Neural Networks 🏆.
If you are interested in similar research directions or would like to explore potential collaborations, feel free to reach out via email. For more details about my background and prior work, please see my CV.
I am visiting Prof. Antoine Bosselut's lab at EPFL in Switzerland to work on MoEs 🇨🇭!
Our work on the Transferability of Machine Unlearning across languages in Multilingual LLMs got accepted at EACL 2026 🇲🇦!
Our Thoughtology paper investigating the reasoning chains-of-thoughts of the Large Reasoning Model DeepSeek-R1 got accepted in TMLR!
I am at NeurIPS and WiML in San Diego to present my work on multilingual MoEs 🇺🇸.
