Food plan generation engine

The problem

Nutrition plans were built by hand. A coach worked from clinical protocols against each client’s conditions, allergies and food habits. Hours per client. Growth was capped at the number of experts that could be hired, and the standard of a plan depended on which coach the client drew.

The system

An engine that generates personalised plans under hard medical constraints. Human verification is calibrated per capability rather than as one blanket policy. Missing safety data fails closed: if a required field is missing or unreadable, generation is blocked and the case routes to a human.

What happened

The time to create one food plan came down by almost 80%. That was the goal: speed, with brand consistency underneath it. Every client gets a plan to the same standard rather than the standard of one coach on one day, and work that scaled with headcount now scales with the platform.

Safety was the thing that could not be traded away to get there. No clinical safety incident.