
Software People Stories · 2026-05-29 · 34 min
My guest today is a good friend and colleague - and not to forget with whom I was a co-author for a book, Paramu Kurumathur. In this episode, Paramu discusses how his recent development work evolved from small Google Apps Script utilities copied and adapted from online examples to building AI-connected applications via APIs to tools like Gemini and ChatGPT, including enabling Q&A over his book content. He describes surprises from “conversing” with his books - especially that LLMs retain details he has forgotten - while noting key risks such as hallucinations and the need for precise prompts. He explains learning Cursor with guidance from our colleague, Raja, discovering that it can generate code, and rapidly producing a proof of concept that maps citizens to the Government welfare schemes using PDFs, Chroma DB, sentence transformers, and queues - work that took about a week instead of months.