RedWings AI

RedWings is a personalized AI sports coach that analyzes an athlete's sports clip using computer vision and biometrics to give real coaching advice for improvement and injury prevention.

Comment

Inspiration

Red Bull is known for pushing the limits of human performance in extreme sports, from snowboarding and motocross to parkour and BMX. That mindset inspired our team at IrvineHacks. Injury rates in extreme sports are high, and most athletes do not have access to professional coaching or biomechanical analysis. Small technique mistakes often go unnoticed until they lead to strain or serious injury. We wanted to build something that makes high quality, data driven feedback accessible to everyday athletes.

What It Does

RedWings AI allows extreme sports athletes to upload a video of themselves performing a trick, drill, or training run. After creating a profile that includes their sport, skill level, body metrics, and training frequency, the platform analyzes their movement using computer vision to extract biomechanical data such as joint angles, symmetry, and movement velocity. That structured data is combined with the athlete’s profile and sent to GPT-5.2 to generate personalized coaching feedback, including form corrections, safety insights, recommended drills, and an overall performance assessment tailored to the individual.

How We Built It

We built a full stack web application using a React and Tailwind CSS frontend with a FastAPI backend. On the backend, we used MediaPipe Pose to detect body landmarks and compute biomechanical metrics such as knee symmetry, hip angles, arm positioning, and movement speed. These metrics are packaged with the athlete profile and passed to GPT-5.2, which generates a structured coaching report. The system runs from raw video upload to AI generated feedback in a single API driven workflow, creating a smooth and cohesive user experience.

Challenges We Ran Into

Our biggest challenge was coordinating four developers building separate systems in parallel. Each team member worked on a different layer, including computer vision, LLM integration, API routing, and frontend UI. When we integrated everything under time pressure, we ran into data mismatches, naming inconsistencies, and merge conflicts. A key technical hurdle was crafting prompts that consistently produced structured, specific, and actionable coaching feedback instead of generic responses.

Accomplishments We’re Proud Of

We are proud that the full pipeline works end to end, transforming a real uploaded video into structured biomechanical data and coherent, personalized coaching advice. We are also proud of the interface, which feels polished and intentional rather than like a typical hackathon prototype. The final product looks and feels like something an athlete would genuinely use to improve their performance and reduce injury risk.

What We Learned

Through this project, we learned how to architect a system so that independent components can be developed in parallel and integrated cleanly. We gained hands on experience with pose landmark detection, structured prompt engineering for reliable JSON outputs from LLMs, handling multipart video uploads between a React frontend and a FastAPI backend, and managing Git workflows within a fast paced team environment.

What’s Next for RedWings AI

Next, we plan to expand support to a wider range of sports and movement types so the platform can adapt to different athletic demands. We also aim to improve the accuracy and robustness of our biomechanical analysis by refining pose detection, incorporating more advanced movement metrics, and making the system more resilient across varied video conditions. Our long term vision is to build a scalable AI coaching platform that delivers precise, data driven performance insights to athletes everywhere.

Built With

Share this project:

Updates