For this Hacktoberfest Weekend Challenge, I built FriendFit, an AI-powered fitness companion designed around a simple idea: Fitness is easier when you have someone keeping you accountable.
I built FriendFit for a friend who wanted to stay consistent with fitness but struggled with motivation, planning, and knowing what to do each day.
Instead of creating another complicated fitness application packed with dozens of features, I focused on a small set of useful experiences that someone would actually use:
The goal was not to replace a professional trainer or medical advice. FriendFit is designed to act more like a personal fitness buddy that helps a friend stay on track.
The project follows the "Build for a Friend" theme by solving a real, everyday problem: staying consistent when you're trying to improve yourself alone.
🎥 Demo Video:
💻 GitHub Repository:
https://github.com/justayushmani/FRIEND-FIT The repository contains the complete frontend and backend implementation of FriendFit.
FriendFit was built as a full-stack web application with AI at its core.
Frontend
Backend
AI Layer
The application is structured around a frontend + backend architecture so that the UI, business logic, and AI functionality remain separated.
I also focused on making the project deployment-ready rather than leaving it as a local prototype.
The AI isn't just a chatbot placed on top of a fitness website.
It is used to make the experience more personalized.
Instead of giving every user the same generic workout experience, FriendFit can use the user's goals and context to provide more relevant guidance.
The idea is:
User → Personal Context → AI Reasoning → Personalized Fitness Guidance
This makes the application feel more like having a fitness companion rather than simply browsing a static workout database.
Open innovation was one of the most important parts of this project.
For a personal application like FriendFit, I don't want the entire experience to depend on one closed AI provider.
Using open-source AI approaches makes it possible to:
This is especially important for a fitness companion because personalization can involve information that users may not want unnecessarily sent to multiple external services.
Open AI ecosystems also make experimentation much easier. Instead of treating the AI model as a black box, developers can build the surrounding system, change the model, experiment with different approaches, and continuously improve the experience.
For FriendFit, open innovation means having control over the technology instead of simply consuming an AI API. I used an AI coding agent during the development process to help build, debug, improve, and prepare FriendFit for deployment.
The development process included:
Agent Session:
[ADD YOUR DEVRELAY / AGENT SESSION LINK HERE]
Building FriendFit in a limited hackathon window taught me that a good AI application does not need hundreds of features.
The biggest challenge was deciding what not to build.
Instead of creating a huge fitness platform, I focused on the core problem:
How can technology make it easier for someone to stay consistent with their fitness goals?
That helped me focus on usability, personalization, and a simple experience rather than feature count.
I also learned how important the surrounding engineering is. The AI model is only one part of an AI application. The frontend, backend, validation, prompts, error handling, deployment, and user experience all have to work together.
I am entering the following partner categories that apply to my implementation:
The project is also automatically eligible for the Overall Winner category because it is a qualifying submission for the challenge.
FriendFit started with a simple question:
"What could I build that would actually help a friend?"
The answer wasn't another massive application.
It was a small, approachable AI companion that helps make fitness feel less lonely and more achievable.
That's what I wanted FriendFit to be:
A friend for your fitness journey. 🤝🏋️
This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend