Fitdyz AI

Problem

Many people want to get fit but do not know how to structure workouts or eat well, and motivation fades fast once the first excitement wears off.

Approach

Users enter their details and the app builds a personalized workout and diet plan, then tracks calories, workouts and meals in one place. FitBot, the in-app assistant, answers questions grounded in each user's own plan through a retrieval pipeline over a vector database. Challenges reward tokens that users spend on live chat with a real coach over authenticated WebSocket rooms, while Redis and BullMQ queues keep heavy AI generation off the request path.

Outcome

A clear plan, steady motivation, and fitness that is easier to stick with. Under the hood it runs like a production system: relational PostgreSQL schema with Prisma, role-aware APIs, async queues and realtime chat.Self-reportedArchitecture as built. No public artifact yet.