Zeno: AI Fitness App Backend with LLMs (Client Project)
Tran Tien Van
Client project. Built by Van Data Team for Zeno, an AI fitness, nutrition and habit app on iOS. Live product: zenoapp.io
The challenge
AI-generated training and nutrition programmes reach a real body, so a low-confidence generation cannot be allowed to ship.
What we built
A NestJS backend on MikroORM over PostgreSQL on AWS RDS
Domain modules per vertical: strength, nutrition, meditation, breathwork and fasting
Multi-provider routing across Claude 3 Haiku, OpenAI and Gemini
Firebase JWT authentication, Pro-tier gating and event-driven processing on AWS SQS
Jest + supertest suites; a Next.js 16 / React 19 demo frontend on Vercel
Results
85% confidence threshold gating every AI-generated programme
Domain modules across strength, nutrition, meditation, breathwork and fasting
Multi-provider routing across Claude 3 Haiku, OpenAI and Gemini
Key decision
Generation is separate from delivery, so a low-confidence result is withheld rather than shown with a caveat.
Stack: NestJS, MikroORM, PostgreSQL, AWS RDS, AWS SQS, Firebase Auth, Claude 3 Haiku, OpenAI, Gemini, Next.js 16, Vercel.
Full case study: vandatateam.com/portfolio
Like this project
Posted Nov 5, 2026
Client project for Zeno: NestJS mobile-app backend with Claude, OpenAI and Gemini routing; AI programs ship only above an 85% confidence gate.
Likes
0
Views
1
Clients
Zeno
