Study Twin | Devpost

Inspiration

The idea for Study Twin was born from the need to empower school students to understand and improve their academic habits through self-comparison, actionable feedback, and teacher support. We noticed that students often struggle to visualize their "ideal study self" and don't get tailored, positive nudges—so we set out to build a digital twin that can guide and motivate them, with real teacher involvement.

What it does

Study Twin is a web + mobile AI tool for school students (Grades 6–12). Each student gets a personalized virtual twin—an ideal version of themselves—generated using simple rule-based AI (heuristics). Study Twin lets students:

Enter their study habits, grades, and self-assessed mastery levels
Instantly generate their virtual twin’s optimal habits, mastery scores, and micro-habit plans
Compare their actual progress against their twin, visually (radar/bar charts)
Take subject quizzes to update their mastery levels and see the gap close over time
Join classrooms with a code, so teachers can track groups, give feedback, and export reports

How we built it

Frontend: React (Vite), styled with Tailwind for a modern, responsive UI on both phone and laptop
Backend: Convex (serverless DB + server functions) for authentication, data, and classroom logic
Data Model: Students, teachers, classrooms, quizzes, and twin profiles all stored as type-safe collections in Convex
Export: Used jspdf/html2pdf.js to allow PDF downloads of the student-vs-twin action plans and progress reports
AI Features: Implemented a heuristic twin generator (rule-based logic, e.g. boost mastery +20, set hours target) and linear grade prediction using:- predicted_grade=0.5×avg_mastery+0.2×study_hours+20\
Mobile/Laptop: Tailwind responsive design for a smooth experience everywhere

Challenges we ran into

Balancing Simplicity and Utility: Keeping the heuristic AI simple enough for a 24-hour MVP, but detailed enough to feel real and useful to students and teachers.
Role-Based UX: Designing flows for both students (self-improvement focus) and teachers (classroom/group management, feedback).
Convex Integration: Figuring out real-time sync and schema-less but typed data modeling sped us up, but learning Convex nuances was a challenge.
Visual Comparisons: Conveying improvement visually (color-coded charts, delta indicators) needed careful chart design.

Accomplishments that we're proud of

Delivered a real-time, cross-device solution.
Implemented smooth student onboarding, classroom joins, and instant twin generation with heuristic logic.
Created a genuinely useful teacher dashboard for group/class tracking.
Enabled exportable PDF reports for both students and teachers.

What we learned

Heuristics are powerful for MVPs—simple rules can simulate “AI” well enough for most hackathon use cases.
Rapid prototyping with Convex lets you ship features without complex infrastructure.
Role-based flows are essential for multi-user educational apps.