I am Ajinkya Kiran Mulay. I work on large-scale LLM safety systems at Meta and pursue independent research on sparse recovery, Transformer optimization, and verifiable AI systems.
I completed my PhD in Electrical and Computer Engineering at Purdue in 2024. That work lived at the intersection of differential privacy, federated learning, and sparse recovery. SPriFed-OMP became the clearest expression of the thesis: recover a small support, communicate less, and still retain a formal account of what the algorithm can guarantee.
More recently, I have been applying the same instincts to optimizers and model analysis. I care about methods that expose their mechanism, artifacts that reviewers can inspect, and experiments that distinguish a promising idea from a confirmed result.
The site is called The Pursuits because OMP, FoBa, and related pursuit algorithms run through much of my technical work. It also describes the broader project: keep following difficult questions without sanding off the uncertainty.