STEMspire

We were inspired by the idea of creating a space where students, especially those in STEM fields, can find mentors whose experiences truly resonate with their goals, backgrounds, and challenges. Many students struggle to find relatable mentors, and we wanted to use technology to make those connections more meaningful and accessible. Our goal was to bridge that gap through an interactive, AI-assisted tool that feels personal and supportive.

Challenges we faced: Technical setup: At first, integrating APIs and handling environment variables was confusing. We ran into authentication and rate limit errors when trying to use live OpenAI embeddings. Data handling: Structuring and formatting our mentors.json file correctly so Streamlit could read and display it took trial and error. Deployment: Setting up Git, connecting to GitHub, and deploying to Streamlit Cloud were new skills that took some patience to troubleshoot. Simplifying AI: We had to find creative, free solutions to mimic AI embeddings once we hit API limits — which taught us how to problem-solve and make the app still work smoothly.

Accomplishments we’re proud of: We built a fully functional interactive web app that recommends mentors based on user input. We learned how to combine programming, data, and design to make a meaningful product. We successfully deployed our app to Streamlit Cloud, creating a live demo link that anyone can use. We overcame multiple technical challenges independently, learning to read errors, debug, and adjust our approach.

What we learned: How to use Python and Streamlit to build web apps from scratch. The basics of machine learning concepts like embeddings and cosine similarity. How to use Git and GitHub for version control and project deployment. How to think like developers — breaking big problems into smaller steps and finding creative solutions. Most importantly, how teamwork and persistence can turn a simple idea into a working, shareable app.