MET

MET is a nutrition and training platform built end to end for a working strength coach's practice. It is invite-only by design: the coach provisions each client, clients log meals and training, and pain flags surface at the top of the coach's intake queue so a human sees risk first.

This is the conventional full-stack case, and I own all of it: a 49-model Postgres schema through Prisma, typed APIs over it, authentication and per-client isolation, the decision engines for progression, scheduling and nutrition, and the React interface the coach and their clients actually use. The domain modeling is the hard part. A training week is not rows in a table, and getting that shape right is what makes the rest of the product simple.

The privacy posture is structural: client media is app-mediated so storage URLs never reach the browser, secrets stay server-side, and the production entry path is invite-only through Clerk. No client sees another client's data, and that is enforced in the data layer rather than in the interface.

Honest status: complete and verified end to end on a production build, held to 319 automated tests; launching with its first real clients now.