This image is currently processing. Please be patient...
When I first got into fitness and weightlifting 5 years ago, I had no idea where to start. With the amount of different exercises, training styles, nutrition fads on the media and misconceptions online, it took me a few years of research and experimenting to get a good grasp on how to create an effective, safe workout program with a healthy, well rounded diet. Recently I became a personal trainer, and want to build a service to make personal training services more accessible to anyone new to fitness.
What it does
Our app personal trainer chatbot talks to you to determine your goals, asks for your physiological measurements and does a fitness assessment to gauge your fitness level. Our AI algorithms then generate a completely personalized workout program that accounts for your past injuries, weaknesses, aiming to improve your fitness overall and help you reach your goals. Using a research supported mathematical model to determine the calories and macronutrients required to reach your goal, designed to compliment your workout plan.
How we built it
We used a variety of APIs and software resources trying to create a clean, smooth and easy to navigate app such as react, node.js, express and firebase. The compuational simulations of our nutritional differential equations were modelled using numpy, and we used an RNN with pytorch that would generate each exercise indivudally and recursively until an entire workout program was complete - trained with a dataset of almost 50,000 workouts.
Challenges we ran into
Ultimately, our biggest difficulty was completing and integrating the ambitious features and services we planned to incorporate in time. What challenged us the most was getting a good enough understanding of the biomedical engineering concepts behind the mathematical model for the diet, in addition to designing an AI architecture capable of effectively delivering the personalized training programs we were aiming for - as well as obtaining a dataset large enough to train it.
Accomplishments that we're proud of
We managed to complete almost the entire front end, most of the backend, a preliminary AI model and decent simulations for the diet within the short amount of time, despite our lack of integration.
What we learned
We learned a lot more about biomedical engineering concepts, front and back end design strategies, and AI architecture challenges.
What's next for Moredovate
We plan on pursuing this project as a startup so we can help other people who shared the same difficulties as a beginner in fitness as I did.