The Problem It Solves
RepWise addresses the lack of personalized, guided, and interactive fitness experiences in conventional workout apps. Most fitness platforms provide static plans, minimal feedback, and fragmented health tracking, making it difficult for users to stay consistent and train correctly.
RepWise solves this by offering:
- Personalized Workout Plans tailored to user goals, fitness level, and available equipment
- Real-time Form Correction using live pose detection with voice feedback to reduce injury risk
- Interactive Workout Sessions with real-time scoring and feedback
- Comprehensive Nutrition Tracking including macro tracking and barcode-based food logging
- Gamified Fitness Experience through PVP challenges and achievement systems
- Unified Health Data Sync across devices using WearOS and Health Connect
Together, these features create a guided, engaging, and data-driven fitness ecosystem that adapts to each user.
Challenges We Ran Into
WebRTC Stability Issues
Implementing real-time video-based form correction introduced major challenges due to unstable network conditions and limited team familiarity with WebRTC. This led us to switch to Agora, which provided a more reliable and scalable real-time communication solution.
OCR Accuracy for Nutrition Labels
Accurately scanning nutrition tables from food packaging proved difficult. Many OCR models failed to handle inconsistent layouts and fonts. Mindee was the closest viable option, though its accuracy was still limited.
AI Model Constraints
Smaller Gemini models struggled with function and tool calling, while larger models introduced latency and compatibility issues with the OpenAI toolspec. Balancing performance, accuracy, and integration compatibility required multiple iterations.