Thestoryso far.

At DRDO I wrote real-time control logic for a military jet engine, where a millisecond of lag isn't a bug — it's a flameout. At Springer I published a deep-learning model that reads chest X-rays well enough for a radiologist to take seriously. Somewhere between those two extremes — software that can't fail and software that has to be believed — is the engineer I'm becoming.

These days I build production AI: LLM-powered applications, retrieval-augmented generation over real document corpora, and automation that quietly deletes hours of manual work. I like owning the whole arc — training the model, then shipping the interface a human actually touches — because the hard, interesting problems usually live in the seam between the two.

I believe the best technology disappears; it just feels inevitable. Whether it's a face-recognition pipeline, a mental-wellness companion, or an autonomous navigation system, I obsess over the small, unglamorous details that make software feel effortless to the person on the other side of the screen.