My work spans the full AI stack at S&P Ratings' commercial division, from model research and architecture to deployment and adoption across teams and geographies. The hardest problems tend to show up when AI capabilities run into the reality of messy financial data.
Before S&P, I spent two years at Praedicat (now part of Moody's) building NLP and deep learning pipelines on scientific literature to surface emerging insurance risks. I did my M.Sc. in Data Science at NYU (2017), where I TA'd for Prof. Yann LeCun and did research with Prof. Kyunghyun Cho on multi-document summarization.
Right now I'm most interested in AI agents and agentic systems, knowledge graphs as structured memory for LLMs, pricing optimization, and multimodal document understanding. Most of what I build at work comes down to making these systems actually hold up when the data is noisy and the decisions matter.
Happy to talk research, AI agents, speaking, or anything in between. If you have a half-formed idea or a concrete project, send me a note.