I ship the parts of AI systems that don't demo well.

I studied computer science, math, and statistics at Christ University, then went deeper into applied computing with an MCA at PES University. Somewhere in between, the math stopped being an abstraction and became the thing that decides whether a retrieval system returns the right six passages or a confident, wrong answer.

At Caasdi Global, a construction-tech startup in Bengaluru, I usually end up owning both ends of a problem: the AI system that reads a tender document, and the pipeline that gets it there reliably. I've built RAG pipelines and agent workflows, and spent just as many hours on the unglamorous parts — feasibility studies, competitive intelligence reports, invoice and payslip generators. Understanding a business well enough to automate its paperwork makes you better at the AI part too.

Outside work I'm building an agent observability platform from scratch — tracing, lineage, evaluation — mostly to understand what actually happens after agent.invoke() returns. I'm also working through the AWS and Databricks data engineering certifications, one gap analysis at a time.