Inspiration

As a university student who's also spent time building a platform for students and job seekers, we kept seeing the same problem: study material piles up fast — dense PDFs, lecture notes, long docs — and there's rarely enough time to turn it into something actually useful for revision. We wanted a tool that takes whatever a student throws at it and instantly turns it into something they can actually study from: summaries, flashcards, and quizzes.

What it does

Study Buddy lets a student upload any study material — a PDF, Word document, or plain text — and instantly generates:

  • A clear summary of the material
  • Auto-generated flashcards (flip to reveal the answer)
  • An auto-generated quiz (multiple choice + true/false) with scoring
  • An "Explain Simply" mode that re-explains any concept in the simplest possible language with an easy analogy

Everything a student uploads is saved to a personal dashboard so they can come back and keep reviewing.

How we built it

Study Buddy is built entirely in plain HTML, CSS, and JavaScript — no frameworks, no backend server. File text is extracted client-side using pdf.js (for PDFs) and mammoth.js (for Word docs). That extracted text is sent to the Gemini API with structured prompts asking for JSON output, which is what powers the summaries, flashcards, and quizzes. All user data (uploads, flashcards, quiz history) is stored locally with localStorage, so the whole app runs and deploys as static files. We used Google Antigravity as our AI dev agent to scaffold and speed up the build.

Challenges we ran into

Getting clean, reliable text out of PDFs and DOCX files in the browser took some tuning, since formatting varies a lot between documents. Getting the AI to consistently return well-structured JSON (so the flashcard and quiz UI never broke) took several rounds of prompt refinement. Doing everything client-side, with no backend, also meant being extra careful about how and where the API key and stored data were handled.

Accomplishments that we're proud of

We built a fully working, AI-powered study tool with zero frameworks and no backend — just HTML, CSS, and JavaScript. Every core feature (summarizer, flashcards, quiz, simplified explanations) works end-to-end, and the app stays fast and lightweight.

What we learned

We learned a lot about prompt engineering for structured/JSON AI outputs, client-side file parsing, and how much you can build without reaching for a framework. Working with an AI dev agent (Antigravity) also taught us how to write clear, structured build prompts to get better results faster.

What's next for Study Buddy

Next up: spaced-repetition review for flashcards, support for more file types (like slides and scanned images via OCR), progress tracking and study streaks, and eventually a lightweight backend so students can sync their study sets across devices.

Built With

Share this project:

Updates