- Ronak Mehta
- Math-ier Machine Learning
- ronakrm @ the big G's mail
My dissertation research focused on methods for efficiently identifying important subsets of features, parameters, and samples in modern ML settings. Current and future interests for me revolve around applying some of these ideas to interpretability and safety, and more broadly exploring issues around alignment. I’m currently working on a couple of projects in the guaranteed-safe and provably-safe AI space, stay tuned!
For a take-away, check out my Résumé/CV.
Recent News
- I participated in the ML and Alignment Scholars Program in Berkeley this summer, and am continuing work on projects in guaranteed-safe AI.
- I started work at Orca, a new startup working on memory augmentation for traditional and language ML models.
- I travelled to Rhodes for ICASSP 2023 to present our work on Robust Blind Deconvolution via Mirror Descent.
- I presented our work on Discrete Optimal Transport (top 25%!) in Kigali, Rwanda at ICLR 2023.
- I participated in a one-month Interpretability Experiment at Redwood Research.