Inspiration
A passion for a science-based approach to lifting is what led us down the path to creating an AI personal coach that can manage schedules. The idea stuck with us because we all agreed that if we were able to fully flesh out the app, we would all use it. You can receive a full-time personal coach in a matter of seconds.
What it does
The program collects information about the user to create a baseline schedule. The program centers around communicating with the user and editing their schedule based on feedback. The AI is catered to address things like health concerns and injury as a priority, and helping the user slowly build up their muscle while achieving either weight gain or weight loss goals.
How we built it
We used the power of Gemini 2.5 Flash to create our AI fitness coach. By tinkering with our prompts, we were able to create accurate and useful exercise information that we could then extract into a Python-based backend. We used React to create a web-based app for the user interface. To bridge the gap between our user interface and our LLM, we used Flask.
Challenges we ran into
We ran into challenges at almost every step. It was our first time using React, so we stumbled through learning it, and it was quite difficult for us to find out how to take information from the AI and use it in such a way to populate our workout calendar. We also use an AI journal to openly communicate with the user, and sometimes it takes a while to fine-tune the prompt to make sure it properly addresses issues, as it often overcompensates for small issues.
Accomplishments that we're proud of
We were able to accomplish our first React project and our first implementation of an LLM in a program. We are also proud of the app idea as there's lots of room for improvement for a genuinely useful app that we could continue development on in the future.
What we learned
We learned how an LLM can be used in programs to enhance a user experience and how to target Gemini with specific prompts to optimize specific user experiences. Some of us also improved some Python basics (some of us were C++ coders), and we all got a great feel for how React works, which can be a useful tool for anyone in the future.
What's next for DynaFit AI
For DynaFit AI, we would like to continue to optimize the user experience by perfecting our Gemini prompts, and we have plenty more features that we brainstormed that we can implement. As of right now, our project runs locally, but we would love to make a public version for all users to have access to.
Log in or sign up for Devpost to join the conversation.