// StudyBridge: Building an AI Study Companion in Public at HACKHAZARDS '26
The Problem
Every student has been there: a stack of PDFs, scattered handwritten notes, and an exam date creeping closer, with no clear plan on how to actually get through it all. Existing study tools usually solve one piece of this puzzle — a notes app here, a flashcard app there, a separate tutor chatbot somewhere else. Nothing ties it together, and almost nothing works well in a mix of Hindi and English, which is how most Indian students actually think and talk while studying.
That's the gap StudyBridge was built to close: a single AI-powered companion that takes whatever study material you give it — a PDF, a photo of handwritten notes, or just pasted text — and turns it into notes, a quiz, a place to ask doubts, and (eventually) a full study plan, all in the language you're comfortable with.
The Idea
The core insight was simple: most of a student's study workflow revolves around one piece of content at a time — a chapter, a set of notes, a topic. So instead of building four disconnected tools, StudyBridge is built around a single upload (or paste) that feeds every other feature:
- Smart Notes — turns raw material into a clean, structured summary with key points highlighted
- Quiz Generator — creates multiple-choice and short-answer questions from the same content
- Doubt-Solving Chat — a chat interface grounded in the uploaded material, so answers stay relevant
- Bilingual Explanations — a toggle between English, Hindi, and Hinglish, since code-switching is how most students actually study
- Study Plan Generator (in progress) — breaks a syllabus down into a day-by-day plan against an exam date
- Voice Answers & Break Reminders (in progress) — small touches aimed at making studying feel less like a grind
Tech Stack
StudyBridge was built almost entirely through Google AI Studio, using a "vibe-coding" approach — describing the app and its features in plain language and iterating on the generated app rather than hand-writing every line from scratch.
- Gemini API (Flash model) for document understanding, notes generation, quiz creation, and chat
- Multimodal input to handle both typed PDFs and photos of handwritten notes
- Streaming responses so notes and answers appear progressively instead of forcing users to stare at a loading spinner
- A simple web front-end scaffolded and refined directly inside AI Studio
Implementation Journey
The first version focused on just one flow: upload a file, get notes back. Once that worked, features were layered on one at a time — quiz generation next, then the doubt-solving chat reusing the same document context, then the bilingual toggle.
A key early decision was adding a "Paste Text" option alongside file upload. PDF parsing can be unpredictable — formatting quirks, scanned pages, or upload errors are common — so giving users a reliable fallback (just paste the text directly) made the app noticeably more robust, and doubled as a safety net for live demos.
Challenges Faced
PDF and paste reliability. The paste-text flow initially threw errors on the first attempt due to state not being captured correctly before the API call fired — a classic race condition between UI state and the request. Fixing this meant tightening input validation and making sure the app only called Gemini once the pasted content was confirmed non-empty.
Response speed. Early versions felt sluggish because the app was using a heavier model than necessary, sending long, repetitive prompts, and waiting for the entire response before showing anything. The fix was threefold: switch to Gemini Flash for speed, trim the prompt down to only what's needed, and stream the response so text appears within seconds instead of after a long wait.
Balancing scope with time. With seven planned features, the biggest challenge wasn't technical — it was discipline. Prioritizing a fully working core (upload/paste → notes → quiz → chat → bilingual toggle) over a half-finished version of everything made for a much stronger, more demo-ready product.
What's Next
The study plan generator, voice answers, and break reminders are the next additions, followed by more testing on real student notes across different subjects and handwriting styles to make sure the core experience holds up outside a controlled demo.
Building StudyBridge for HACKHAZARDS '26 was a good reminder that the best AI tools don't try to do everything for everyone — they take one real, everyday frustration and quietly make it disappear.
Built for HACKHAZARDS '26 by Namespace Community. #HACKHAZARDS26 #GeminiAI #GoogleAIStudio
Top comments (0)