Studylane-AI study companion

Turn every study resource into a learning plan that knows what you know and helps you master what comes next

Comment 2

Inspiration

College and high school students frequently study from multiple fragmented sources: slide decks, PDF textbooks, research papers, handwritten notes, and YouTube lectures.

Switching back and forth between an hour-long video, a 60-slide presentation, and dense textbook pages creates unnecessary friction. Most existing AI study tools act like disconnected chat windows. They generate summaries that are difficult to verify, hallucinate details, and provide no clear way to trace facts back to the original source.

We created Studylane to solve this issue. It is a multimodal academic workspace that ingests any study material, extracts the core concepts, and provides dynamic citations that link directly to the exact page, slide, or video timestamp in a side-by-side preview panel.

What it does

Studylane converts scattered learning materials into an organized, verifiable study system:

Multimodal Document and Video Ingestion

  • Ingests PDFs, PowerPoint presentations, Word documents, diagram images, and YouTube lecture links.
  • Uses Gemini to review video content, confirm that the material is instructional, and produce timestamped transcripts.

Dynamic Citations and Side-by-Side Preview Panel

  • Generates clickable citation tags such as [Page 4], [Slide 12], or [05:20] for notes, flashcards, and tutor answers.
  • Clicking any citation opens the side-by-side document preview panel:
    • For PDFs and slide decks, it displays the original file inline and navigates straight to the cited page.
    • For YouTube videos, it embeds the player, automatically jumps to the cited timestamp, and highlights the corresponding line in the transcript.

Study Studio

  • Smart Notes: Produces comprehensive study guides, quick revision sheets, and exam cram notes with LaTeX math equations and inline source citations.
  • Active Recall Flashcards: Provides interactive flashcards with confidence ratings and exact source references on the back.
  • Diagnostic Quizzes: Generates multiple-choice and short-answer questions with automatic scoring, answer explanations, and citation links.
  • Knowledge Graph: Visualizes key concepts and their connections across documents as an interactive 2D node graph.

Grounded AI Academic Tutor

  • A conversational tutor that answers questions using the uploaded materials as its primary reference.
  • Offers three comprehension levels: Simple (intuitive analogies), Intermediate (college introductory), and University (rigorous academic depth).
  • Prioritizes source documents while providing explanations for missing concepts rather than stopping with false limitations.

How we built it

Multimodal AI Engine

  • Built with Google Gemini using the google-genai SDK for multimodal video inspection, OCR image understanding, and structured data generation.
  • Directly processes public YouTube videos to produce accurate timestamped transcript segments and assess academic substance.

Backend Infrastructure

  • Developed using Python and FastAPI as an asynchronous REST API.
  • Implemented file processing pipelines with pypdf, python-pptx, python-docx, Pillow, and youtube-transcript-api.
  • Uses SQLite for persistent storage of materials, quizzes, flashcard progress, and chat sessions.
  • Caches original uploaded files so they can be viewed inline in the browser.

Frontend Interface

  • Built with React and Vite for a fast single-page user interface.
  • Utilizes react-force-graph-2d for the interactive knowledge map.
  • Uses react-markdown and KaTeX for rendering markdown notes and mathematical formulas.
  • Icons provided by lucide-react.

Challenges we ran into

  • Preserving Original Document Layout: Extracting raw text from PDFs and slide decks stripped out diagrams, formulas, and visual context. We addressed this by preserving original binary files and implementing an inline document viewer that opens the original PDF to the specific page referenced.
  • Synchronizing Video Timestamps: Parsing dynamic citations from AI responses and syncing them with an embedded YouTube player required regex pattern detection, time conversion, and automated playback seeking.
  • Validating Educational Content: Many online videos have academic-sounding titles but contain entertainment or irrelevant commentary. We created an evaluation pipeline that inspects the actual instructional content before ingestion.
  • Balancing Accuracy and Helpfulness: Overly strict grounding prompts often led the model to reply that it could not answer if a detail was absent from the uploaded text. We balanced the system prompt so it prioritizes the document while using broader academic knowledge to explain and bridge gaps.

Accomplishments that we're proud of

  • Seamless Source Verification: Students can click any citation pill to verify an answer against the original slide, textbook page, or lecture video moment.
  • Unified Workspace: Students can combine different formats like a PDF textbook, a slide deck, and a recorded lecture into one study workspace.
  • Simple Local Setup: Created self-contained setup and startup scripts that configure virtual environments, install dependencies, and launch both frontend and backend servers automatically.

What we learned

  • How to utilize Gemini for multimodal video analysis and structured JSON generation.
  • How to structure grounding prompts that maintain strict academic accuracy while preventing unhelpful refusal responses.
  • The value of split-screen layouts in educational software to keep students focused on primary source materials.

What's next for Studylane-AI study companion

  • Voice Tutor Mode: Adding real-time spoken conversations with the Gemini Live API so students can practice verbal recall and oral exam prep.
  • Study Groups: Introducing shared workspaces where peers can collaborate and take practice quizzes together.
  • Handwritten Problem Review: Allowing students to upload pictures of handwritten math or engineering solutions for step-by-step checking.

Built With

Share this project:

Updates

Submission history