The Problem I Wanted to Solve
Students can have all their course materials and still find it hard to study effectively. Notes are scattered across files, it takes time to find a clear explanation, and planning what to study next can feel like a separate task. I wanted to build something that helps students move from “I have the material” to “I know what to learn and how to practice it.”
That idea became ShikhonSathi AI, a student-focused B2C learning companion built for the Ostad Presents YUNet AIgnite Buildathon 2026. Its goal is simple: make studying more organized, useful, and accessible by bringing common learning tasks into one web app.
What ShikhonSathi AI Does
Students can add their own course materials and ask questions about them. The AI is designed to base its answers on the selected material and show its source, helping students understand where an explanation comes from. When a question falls outside that material, the experience is designed to make that clear instead of presenting every answer as though it came from the notes.
The same material can support the next steps in a student’s study routine: generating a study plan, checking off completed study days, and creating practice quizzes with a chosen question count and time limit. Students can search YouTube for topic-related learning resources and save useful answers to revisit later. The interface supports both Bangla and English, so students can study in the language that suits them.
How I Built It
I built ShikhonSathi AI as a full-stack web application with Next.js. Gemini powers the AI study features, while Firebase Authentication and Firestore support Google sign-in and account-based data storage. The app connects the student’s material, AI requests, and saved learning activity into one workflow.
A key design decision was to make studying begin with the student’s own material. Students should be able to add a course resource, ask about it, and continue into planning and practice without having to rebuild their context for each feature.
Challenges and What I Learned
One of the biggest challenges was making the complete experience work together—not just getting an AI response. The material upload, AI request, authentication, saved data, and deployed app all needed to cooperate. I also worked through how to make answers feel relevant to the selected notes, how to signal when a question is outside their scope, and how to make saved study activity available again when a student returns.
This project taught me that a useful AI product needs more than a model call. It needs a real user problem, thoughtful workflows, clear feedback, dependable data handling, and an interface people can use comfortably. I built ShikhonSathi AI around that idea: AI should help students make progress with the materials they already have.
Built With
- cloud-firestore
- firebase
- firebase-authentication
- google-gemini-api
- next.js
- react
- tailwind-css
- typescript
- vercel