
Overview
FitNOVA is a full-stack web application built to personalize health and wellness journeys using advanced AI models. Designed with a user-first mindset, it delivers individualized workout and diet plans based on real user data
—like age, goals, fitness level, and dietary restrictions. With real-time tracking
, adaptive feedback loops
, and rich analytics
, the platform empowers users to stay consistent, stay motivated, and make smarter decisions about their fitness journey.
The project was developed as part of the Agile Software Engineering & DevOps coursework at RV University.
Key Objectives
- Personalized Planning: Generate AI-driven workout and diet plans based on user metrics.
- Progress Tracking: Enable users to monitor their
calorie intake
,workout completions
, andfitness stats
via intuitive dashboards. - AI Integration: Use
Google’s Gemini
andFlash 2.0 Lite
models to generate structured plans tailored to individual needs. - Responsive UX: Design a fast and
mobile-friendly
interface with intuitive user flows. - Secure & Scalable: Ensure robust
authentication
,data privacy
, and seamless performance under real-time workloads.
AI-Driven Innovation
Unlike traditional apps, this project uses prompt-engineered
AI models to create fitness routines and meal plans in structured JSON
. These responses are parsed and rendered on the frontend, enabling editable
, adaptive
, and context-aware
plans.

Technologies Used
Frontend
- Next.js 14 (React 19) - App router with hybrid SSR
- TypeScript - Type safety throughout the stack
- Tailwind CSS + Shadcn/ui - Utility-first styling with pre-built accessible components
- Chart.js - Dynamic data visualization for progress tracking
Backend
- Supabase (PostgreSQL) - Managed, scalable database
- Prisma ORM - Schema migrations and type-safe database queries
- NextAuth.js - OAuth + Credentials-based authentication
- Google Gemini APIs - Real-time AI-powered workout and diet generation
- Serverless Functions - Hosted via Vercel for scalable APIs
DevOps
- GitHub Actions - CI/CD with testing, linting, and Docker builds
- Vercel - Seamless deployment for frontend + API
Development Approach
Agile Methodology
- Sprint Planning: Defined user stories and deliverables for each sprint.
- Daily Standups: Tracked blockers, task updates, and priorities.
- Retrospectives: Iterated based on feedback and user testing.
- CI/CD: GitHub Actions ensured builds, linting, type-checking, and deployment validations.
Core Functional Modules
- User Authentication – Supports
Google
,Discord
,GitHub
,LinkedIn
and Email/Password logins.

- Onboarding System – Multi-step wizard to collect user health metrics and fitness goals.

- AI-Powered Plan Generator – Uses structured prompts with Gemini APIs to return customized workout routines and diet plans.
- Interactive Dashboard – Visualizes calorie stats, completion history, and editable plans.
- Feedback-Driven AI – Captures user modifications and adherence data to refine future plans.
- Test Coverage – >
90% coverage
across75+ automated tests
(unit, integration, and API).

- Validation – Zod schemas validate all forms both on client and server.
- E2E Workflows – Simulated full user journeys from login to dashboard updates.
- CI Workflows – Automatic test runs, coverage reports, and deployments on PRs.

Database Design
The database schema was modeled using Prisma
, leveraging relational schemas
to capture user profiles, workouts, meals, metrics, and AI-generated plans with proper associations and constraints.
You can view the full schema design here .
Challenges Faced
- AI Model Tuning - Prompt engineering for structured and accurate outputs required rigorous testing.
- Data Privacy - Ensuring GDPR compliance with
Supabase RLS
,JWT sessions
, andencrypted storage
. - Plan Adaptability - Implementing feedback loops that actually influence AI outputs and UX.
- Tech Stack Friction - React 19 peer dependencies and
SSR issues
during build pipelines.
Conclusion
This project is more than just a fitness app—it's a data-driven, AI-personalized wellness companion. From secure authentication to intelligent recommendations, every component was engineered with precision, usability, and scalability in mind. It showcases how modern web tech and responsible AI can converge to create highly impactful, real-world applications.