Every project on this site is a prototype I built to work out a problem — how a workflow actually fits together, where AI earns its keep, what a better UI would feel like. Some shipped. Some changed my mind. All of them are here because the act of building them was useful.
I reach for code the way other people reach for a whiteboard. The fastest way I know to check whether a mental model holds is to make it clickable. So most of what's here started as a one-evening sketch to answer a specific question — would this layout really be usable? What happens to the data at that step? Is the friction where I think it is?
I care about the loop more than the artefact — the deck, the diagram, the wireframe are outputs. The thing I'm trying to design is the cycle that produced them: question → prototype → what it reveals → next question. If the loop is running, the artefacts get better on their own.
Mostly I think in prompts. I describe the behaviour I want — a single-file prototype, a live dashboard, a scraper that runs overnight — and iterate with AI until the shape is right. Under the hood it tends to come out as HTML, CSS, and vanilla JS for one-file sketches, React when state gets interesting, and a Python pipeline when something needs to run on a schedule. Figma when I'm working with someone else. The code ships; the prompts are where most of the work actually happens.
Email is fastest: abhirup.ie@gmail.com. If you've landed here from a conversation, a cold intro, or a link someone sent, tell me which — I'll match the reply to the context.
I'm happy to chat about product, prototypes, GTM for AI, or anything at the seam. A good question gets a same-day reply.