Hi, I'm Bec. I design systems and ship the software that runs them.

Freelance quotes usually mean a PDF attached to an email, then silence. Clients wonder where their build stands, and every proposal tool on the market wants a monthly fee to rent you one slice of the answer on somebody else’s domain.

So I built the whole workflow into this site: a login-gated client portal where a quote is a living page. Magic-link sign-in, revision requests that land in my inbox, click-wrap acceptance that emails both sides a generated agreement PDF, and then a live build tracker in the house vocabulary: on the hard, sea trials, launched. Stripe runs the money: deposit on accept, balance before launch, care plans as subscriptions.

Key calls: Postgres row-level security is the boundary, not the API: clients can only ever read their own quote because the database enforces it, so an endpoint bug can’t leak what the database won’t serve. Every client write goes through a SECURITY DEFINER function that re-verifies identity and quote state. And acceptance freezes the full quote (items, prices, terms) into an immutable snapshot, so “what did I agree to” has exactly one answer, forever.

The shop ran on Apple Notes and shouting across the floor. Jobs got lost between the counter and the bay, and nobody could answer “where’s my cylinder?” without walking the building.

Full-stack iOS workflow app with six-stage job tracking, device-based roles (Front Counter, Tech, Admin iPads), and a Supabase backend tuned for offline resiliency and flaky Starlink. It even generates its own QR codes, and the print layouts to sheet them, for asset tracking across the floor.

Key calls: Device-based roles instead of user logins (the shop floor shares iPads and nobody’s typing a password with greasy hands, so identity lives on the device, not the person). Offline-first data layer because the shop runs on flaky Starlink, so the app stays functional even when the connection wants to take the day off. And it generates its own QR codes and the print-sheet layouts to produce them, so asset tracking didn’t need a second tool. Built to survive a real shop, not a demo.

Every time we talked about going fishing, the same interrogation started. “When is high tide?” “Wind speed?” “Is it rough out?” And then the real one, “How do the fish feel about all of it?” No single app answered the actual question: should we go fishing?

So I built one. Scores every hour 0 to 100 for fishing quality by weighing eight environmental factors against per-species temperature profiles for six Gulf species. Tide data from WorldTides, weather and marine conditions from Open-Meteo. Because when I’m not doing this, chances are, I’m fishing.

Key calls: The forecast is an eight-factor weighted scoring model with water temperature as a hard gate: if the temp is wrong for the species, no amount of perfect tide and wind can rescue the score, because that’s how fish actually work. And the whole engine is extracted into FishingForecastKit, a reusable Swift package, so the next water app gets the forecasting brain for free instead of a copy-paste.