EduScan AI

Inspiration## Inspiration

Students constantly juggle textbooks, class notes, and last-minute doubts before exams — often with no one around to explain a tricky concept at 11 PM. I wanted a tool that could look at any page of a textbook or handwritten note and instantly become a tutor, without needing the content typed out first.

What it does

EduScan AI lets a student upload a photo of a textbook page, notes, or a problem set, and choose how they want help with it:

• Explain – breaks down the concept in simple terms
• Solve – works through problems step-by-step
• Summarize – condenses long passages into key points
• Quiz Me – generates practice questions from the content
• Translate – converts the material into another language
• Custom – lets the student ask their own specific question about the image

How I built it

I built EduScan AI as a solo developer using vanilla JavaScript, HTML, and CSS for the frontend, with the Google Gemini 2.0 Flash API handling all multimodal reasoning. Images are sent directly to Gemini with mode-specific prompts, and responses are rendered back into a clean, mobile-friendly chat-style interface. Each learning mode uses a tailored prompt structure to get the most useful output from the model — for example, "Solve" mode explicitly requests step-by-step reasoning, while "Quiz Me" requests structured question-answer pairs.

Challenges I ran into

• Prompt-engineering six distinct modes so each felt genuinely different in tone and usefulness, rather than just re-labeling the same explanation.
• Handling variable image quality (blurry phone photos, skewed angles, handwriting) while keeping Gemini's responses accurate.
• Balancing response latency with a smooth, responsive UI since users expect quick feedback while studying.

What I learned

Working solo end-to-end on this project taught me a lot about designing effective multimodal prompts for Gemini, structuring a UI around distinct "modes" of interaction, and iterating quickly under time pressure to ship a polished, usable product.

What's next for EduScan AI

• Support for multi-page/multi-image uploads (e.g., a full chapter at once)
• Persistent history so students can revisit past scans
• Offline/low-bandwidth mode for accessibility in low-connectivity areas

Built With

• css3
• gemini-2.5-flash
• google-gemini
• html5
• javascript