Exam Rescue — From Exam Chaos to an Exam-Ready Plan

Priya — Sun, 04 Oct 2026 17:02:36 +0000

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend

What I Built

I built Exam Rescue, a student-focused study planning tool for my roommate.

I built it because she often starts panicking when exams are close and doesn't know where to start when there is a lot of syllabus and study material to cover.

Exam Rescue turns that exam-time chaos into a structured workflow.

Students can enter:

Their syllabus
Study materials
Previous year questions (PYQs)
Number of days remaining before the exam

The app then analyzes the inputs, identifies important topics from the PYQs, matches topics with study materials, assigns priorities, and generates a day-by-day study plan.

It also includes an AI Study Coach that gives additional study priorities based on the student's subject, syllabus and PYQs.

The workflow is:

Input → Analyze → Prioritize → Study Plan → AI Guidance

Demo

Live Demo: https://exam-rescue-omega.vercel.app/

The project runs directly in the browser. The AI Study Coach downloads its model when it is used for the first time and performs inference in the browser.

Code

GitHub Repository: https://github.com/priyaupadhyay-2311/Exam-rescue

The project is built as a frontend application using HTML, CSS and JavaScript.

How I Built It

I built Exam Rescue with HTML, CSS and JavaScript, with the study-planning logic running directly in the browser.

For the AI component, I used Hugging Face Transformers.js with the open-weight SmolLM2-135M-Instruct-ONNX-MHA model.

The model runs in the browser through ONNX instead of requiring a separate proprietary AI API backend.

The AI Study Coach uses the student's subject, syllabus and previous year questions to generate short study priorities and advice.

This makes the AI component lightweight and keeps the inference experience directly in the browser.

Why Does Open Innovation Matter?

Open AI technology made it possible to add an AI Study Coach without building the project around a paid proprietary AI API.

Because inference runs in the browser, the project does not need to send every AI prompt to a proprietary AI service.

It also gives the project flexibility: the open model can potentially be replaced or upgraded without redesigning the entire application around a closed provider.