I don't just spec products.
I ship them.

The problem. Small landlords managing 2–15 units track rent in notebooks and WhatsApp — they forget due dates, lose expense records at tax time, and have no single view of who's paid. Existing software is built for large property firms: too heavy, too expensive, and assumes a Western rental model.

My role. Solo, end to end — product, design, and engineering. The interesting part wasn't the code; it was the product decisions:

• Build vs buy: bought reliability (Razorpay, Supabase, auth) so I could spend my time on the product surface that's actually differentiated.
• Data trust as a requirement, not a detail: hardened row-level security so the database itself guarantees no landlord can ever read another's data — one leak would kill trust permanently.
• Sequenced by job-to-be-done: built Properties → Tenants → tracking → Dashboard first (who owes what), then reminders; deliberately cut a tenant-facing app to protect the core.

Where it is. Newly launched and in production — payments and secure multi-tenant data working end to end, with its first landlord actively using it. The goal was to prove I could take a product from problem to secure, paid, deployed software. Next phase: acquisition.

An AI tool built on the Claude API that drafts structured PRDs — problem, goals, users, requirements, and success metrics — from a short product idea. I chose a foundation-model API over fine-tuning for capability-per-dollar and speed to ship, and designed it as assistive, not authoritative: it produces a strong draft a PM edits before trusting, which makes the occasional weak output a non-event instead of a failure.

A product tool, built by someone who thinks like a PM.