About

Hi, I’m Nanda Kumar.

I’m an AI engineer. Right now I’m building a platform that lets enterprise teams ship their own AI agents without writing code — think Lovable or Vercel, but pointed at internal workflows instead of marketing sites. The thesis is that the people closest to a workflow should be the ones shaping the agent that runs it; my job is to make that path short.

Outside of work I spend time on multi-step agentic workflows and context engineering — the unglamorous mechanics of getting the right information in front of a model at the right time. I’m also slowly working back through fundamentals on the principle that the abstractions I lean on every day are only as load-bearing as my understanding of what’s underneath.

I recently finished my master’s at Concordia University. My thesis was on structured information extraction from LLMs — preprint on arXiv. Before that I was a Machine Learning Engineer at Femtherapeutics.

This blog is a working journal: notes from projects, things I changed my mind about, occasional live demos. The point isn’t polish — it’s leaving a trail I can read back.

When I’m not at a keyboard I skateboard, game, and shoot a little photography.

You can reach me at nanda [dot] kumark [at] mail [dot] concordia [dot] ca.