Health and fitness · AI Agents
Zeno Fitness and Habit Platform
Built by Van Data Team for Zeno
Production mobile-app backend and demo frontend for an AI fitness, nutrition, and habit platform, with generated programmes gated on a confidence threshold.
- confidence gate on every generated program
- 85%





Measured outcomes
Outcome 01
85% confidence threshold gating AI-generated programmes
Outcome 02
Domain modules per vertical across strength, nutrition, meditation, breathwork, and fasting
Outcome 03
Multi-provider routing across Claude 3 Haiku, OpenAI, and Gemini
Challenge pressure
AI-generated training and nutrition programmes reach a real body, so a low-confidence generation cannot be allowed to ship.
The situation
Progressive-overload programming and calorie targets are advice a person acts on physically. A general-purpose generator is acceptable when the output is a draft; it is not when the output is next week of training.
What we built
A NestJS backend on MikroORM over PostgreSQL on AWS RDS, with domain-driven modules per vertical covering strength, nutrition, meditation, breathwork, and fasting. Multi-provider routing spans Claude 3 Haiku, OpenAI, and Gemini. Firebase JWT handles authentication, Pro-tier gating controls access, AWS SQS carries event-driven processing, and Jest with supertest covers the suites. The demo frontend is Next.js 16 and React 19 on Tailwind 4, deployed to Vercel.
What changed
Generated programmes are held behind an 85% confidence threshold before they reach a user.
Constraints and decisions
The confidence gate is the product decision that shaped the architecture: generation had to be separable from delivery so a low-confidence result could be withheld rather than shown with a caveat. Routing across three providers exists to keep that threshold reachable at an acceptable cost per generation.
Technology stack
Need a similar system?
If this case study maps to your current workflow, the next useful step is a scoped project conversation with constraints, timeline, and integration needs.
Related case studies
More work like this
Precision medicine
An Azure platform where engines analyse the labs and AI explains them
Rule-based engines produce the analysis; the AI explains it and cannot change it.
Client projectLiveSmall business operations
A supervisor agent that runs 70+ business tools for small-business owners
70+
MCP tool integrations under one supervisor

