A generative AI-powered application that personalizes fitness engagement content based on user profiles, community context, and real-time feedback — deployed using Flask and Streamlit.
This diagram outlines the modular architecture for user profiling, community assignment, and generative post creation using fine-tuned T5/GPT models. Each module is loosely coupled, allowing independent optimization and feedback integration.
- Extracts features like age, fitness goal, experience level, and workout preferences.
- Uses GPT-4 (zero-shot) or BERT embeddings + scoring rules to match users to communities.
- T5 fine-tuned on community fitness posts; generates daily challenges, reminders, and motivational messages.
- Planned feedback integration for post refinement and re-ranking using user engagement.
- Smart user-to-community mapping based on contextual cues.
- Daily motivational content tailored to group goals.
- Live UI using Streamlit; easily customizable via Flask APIs.
- Modular components for easy retraining or upgrading (e.g., swap T5 with LLaMA).
| Category | Tools/Tech |
|---|---|
| Language | Python |
| NLP Models | GPT-4, BERT, T5 (fine-tuned) |
| ML Tools | PyTorch, Scikit-learn |
| Backend | Flask |
| Frontend | Streamlit |
| Storage | JSON, CSV |
| DevOps | GitHub Actions (optional), Docker |
Input:
{
"age": 22,
"goal": "muscle gain",
"experience": "beginner",
"preference": "gym workouts"
}