I build products end-to-end — from the data pipeline to the pixel.

Shipping AI-heavy web products across fintech, healthcare, and developer education. I own the whole stack — the data model, the backend, the interface — and the AI systems that keep the data honest.

Systems that stay honest

MatchMyCard runs a monthly pipeline that scrapes UAE bank sites and pulls every cashback rate, offer, and benefit into structured data — each figure traceable to the bank's own page. Cards get quietly devalued and no one re-reads the fine print; this keeps what users see accurate and provable.

AI as infrastructure

LLM extraction turns messy inputs — bank PDFs, handwritten dental slips — into structured data behind validation gates. At ClinicKhata a correction loop learns each clinic's handwriting, so extraction gets more accurate the more it's used.

One engineer, a fleet of agents

I architect and review, then direct AI agents on the work — coding agents on implementation, and at BehindScale a fleet that researches, drafts, and reviews. The job is the guardrails and checks that keep them on-goal and off slop, so one engineer can run several products at once.

I spent twelve years making shared UI trustworthy at scale — founding a company-wide design system at Informatica, then leading frontend platform work at Agoda — before turning that same obsession with correctness toward products of my own.

What I care about is consistent across all of it: systems where every number can be traced to a source, AI doing real infrastructural work instead of demos, and one engineer — with the right build loop — running several serious products without cutting the corners that matter. I review the parts that count and let the agents handle the rest.