Designing an AI fitness coach that keeps people consistent

Rayafit is an Android fitness app with an AI coach, Raya, who builds a workout and Indian diet plan around your goal, then adjusts it every Sunday based on what you actually did.

I designed it end to end as my own product in one month: competitor review, PRD, user flow, brand, UI and design system. The Android MVP is now in development.

Most people don't quit fitness apps because the workouts are too hard. They quit when a busy week breaks the plan and the app makes them feel they've failed.

So the challenge was to design for the bad week as much as the good one: realistic plans, familiar food, and AI changes people can trust because every one of them is explained.

A review of HealthifyMe, Cult.fit and Fitbod showed the gap: no app adapts both workouts and meals to what you did last week. I organised Rayafit around three outcomes (train, eat, stay consistent) and five tabs.

Today shows one workout and one Start button. Meals are a timeline built around the workout, in dishes Indian homes already cook. Every Sunday, Raya shows each change with its reason, and nothing applies until the user agrees.

The AI only picks from a reviewed exercise and meal library, and rules check safety limits, calories and protein before any plan is shown.

27 Play-ready screens in light and night mode, a full PRD with AI safety rules, a 16-area user flow and a design system with spacing tokens.

Success will be judged after the closed test on weekly consistency, day-30 retention and how often a low day ends in 5 minutes instead of a skip.