INTERNZ
Shikshkasathi-AI in Education
"From Local Host To Global Impact"
Submitted
May 22, 2026, 4:57 AM
Last Updated
May 22, 2026, 5:16 AM
Project Links
Official Winner
This project was recognized for outstanding achievement in the Education category at Catalyst 2K26.
Problem Statement
we have picked this topic because India’s overall pupil-teacher ratio (PTR) is very bad , and grading manually by teachers is a very slow . Which adds up the existing problem , So we have thought that if we can solve both the problems , Our application not just help teachers grade their students faster but also help them to get a thorough review on their mistakes and suggest a perfect study lesson and plan , So now the teacher will focus more on learning and teaching rather than spending all their time grading students. It also saves the cost per teacher which is beneficial for the institutions . We believe that our project that actually be impactful to the society.
Solution & AI Usage
AI Models / APIs Used Google Gemini 2.5 Flash API Used for: Handwriting understanding Answer evaluation Rubric-based grading Feedback generation Lesson plan generation Why Gemini? Strong multimodal capability → reads handwritten answer sheets accurately Fast response time → grading in ~10–25 seconds Cost-efficient for hackathon-scale deployment Good multilingual support → Bengali, Hindi, English feedback Large context window → handles full answer scripts + rubrics together Other Technologies Vercel Next.js 14 + React + TypeScript → frontend & backend MongoDB MongoDB Atlas → database storage JWT + bcrypt authentication → secure login system Tailwind CSS + shadcn/ui → responsive modern UI The main idea was to combine AI grading + teacher workflow automation + multilingual student feedback into one platform.
Full Description
ShikshakSathi is an AI-powered grading platform built for Indian teachers and students to simplify handwritten answer-sheet evaluation. Teachers upload answer keys and student answer sheets, and the system uses Google Gemini AI to read handwriting, grade answers using rubrics, generate bilingual feedback (English + Bengali/Hindi), and create lesson plans based on student mistakes. The platform includes teacher dashboards, student portals, analytics, and manual score overrides to ensure teachers remain in control. It was built using Vercel Next.js, React, TypeScript, Tailwind CSS, and MongoDB MongoDB Atlas. Major learnings included the importance of prompt engineering, UX design, multilingual accessibility, and balancing AI automation with human oversight. Key challenges included handling poor handwriting, maintaining structured AI outputs, optimizing performance for large uploads, and building teacher trust in AI-assisted grading.
Tech Stack
Screenshots
5 attached
