Flexi

Inspiration

The idea for Flexi stemmed from a desire to address the growing health crisis related to poor posture and inactivity. With the rise of sedentary lifestyles, especially in desk-bound jobs, we noticed an increasing number of people suffering from bad posture and its negative effects. We wanted to create something that would help people correct their posture and improve their overall fitness in a personalized way, without the need for expensive trainers or gym memberships. The rise of AI technology gave us the perfect tool to make fitness more accessible and tailored to each individual.

What it does

Flexi uses cutting-edge AI to analyze users' workout videos, providing instant, personalized feedback on posture and form. Users simply record themselves exercising, and our AI evaluates their movements, offering suggestions for improvement. Whether you're a beginner or a seasoned athlete, Flexi adapts to your fitness level and gives you the insights you need to exercise smarter and reduce the risk of injury. It’s like having a personal trainer in your pocket, guiding you every step of the way.

How we built it

We built the backend using Flask, with Fetch.AI's uAgents at the core of our decentralized multi-agent system, which is a key innovation of this project. For intelligent retrieval, we integrated Gemini and RAG powered by Chroma. OpenCV was used to process videos, extract key frames, and compress file sizes for improved efficiency. Additionally, we developed an exercise knowledge base using Chroma, enhancing the RAG system's capabilities. To enable voice interaction, we integrated Deepgram for speech recognition and user feedback.

Challenges we ran into

One of the key challenges we faced was designing a robust decentralized multi-agent system with Fetch.AI’s uAgents, ensuring seamless communication between agents while maintaining system efficiency. Training the AI to provide accurate feedback across various body types and exercise forms also required extensive testing and optimization. Additionally, processing large video files in real-time using OpenCV, while keeping performance smooth, posed a significant technical hurdle. Finally, integrating diverse technologies like RAG, Chroma, and Deepgram into a cohesive user experience without compromising functionality was a complex task that required careful coordination.

Accomplishments that we're proud of

We're incredibly proud of how Flexi can adapt to different users' needs, providing personalized feedback that feels like having a real trainer. The accuracy of our AI in detecting poor posture and giving relevant, easy-to-follow recommendations is a huge achievement. We’ve managed to build a system that not only helps users improve their form but also contributes to long-term health benefits, which was a key goal from the start. Seeing the app come to life and provide real value to users has been incredibly rewarding.

What we learned

Throughout the development of Flexi, we learned a lot about balancing advanced technology with simplicity in design. It was important to make sure the AI's capabilities were not only powerful but also accessible and understandable to users. We also gained valuable insights into human movement and posture correction, which helped us refine the feedback provided by the app. Moreover, managing and processing video data in real-time while maintaining a seamless user experience taught us a great deal about optimizing both the backend and frontend systems.

What's next for Flexi

Next, we’re looking to expand Flexi’s capabilities by adding more types of exercises and integrating deeper AI insights. We want to personalize the fitness journey even further by incorporating features like long-term progress tracking, customizable workout plans, and even nutrition advice. Additionally, we plan to explore partnerships with fitness trainers and health professionals to provide more comprehensive support to our users. Ultimately, our vision for Flexi is to become a full-fledged, AI-powered fitness assistant that not only helps with posture correction but also motivates users to achieve their overall fitness goals in a fun and engaging way.