Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
Inspiration
As a social sciences teacher and doctoral researcher in Educational Sciences, I saw first-hand how neurodivergent students and students with learning difficulties often struggle in math and reading because standard tests are one-size-fits-all. I wanted to build a tool that meets each student where they are instead of forcing them through a fixed curriculum.
What it does
Proyecto Gemini gives students a math and reading assessment. As soon as they log in, the platform starts the test and connects it to an AI engine powered by Gemini. Based on each answer — correct or incorrect — the system adjusts the difficulty and focus of the next questions in real time, building a live picture of the student's level and steadily steering the questions toward what they actually need to practice.
How we built it
The frontend is built in React and deployed on Vercel. Supabase handles the database and authentication, storing each student's session and response history. The adaptive engine calls the Gemini API to generate and calibrate follow-up questions based on the student's answer pattern.
Challenges we ran into
Designing an adaptive algorithm that adjusts difficulty smoothly, without overwhelming or under-challenging the student, took a lot of iteration — especially for students with learning difficulties, where the margin between "just right" and "too much" is narrow.
Accomplishments that we're proud of
Getting a working, live adaptive assessment flow for both math and reading, deployed at proyecto-gemini-phi.vercel.app.
What we learned
Real-time adaptive difficulty adjustment for neurodivergent learners needs to weigh both accuracy and response patterns, not just right/wrong answers.
What's next for Proyecto Gemini
Expand to more subjects, refine the adaptive model with more student data, and pilot the platform with more students with learning difficulties.
Log in or sign up for Devpost to join the conversation.