CiteTrace

Intuitively visualize, interpret, and connect hundreds of pages of research in minutes.

Comment 1

Inspiration

As student researchers we’re familiar with how messy the research process is. Countless dense pages of academic content makes understanding the whole picture a daunting process. CiteTrace streamlines this process, allowing people to visually understand how the ideas, methods, and various aspects of research relate to one another across their entire project.

What it does

CiteTrace transforms your collection of academic works (papers, textbooks, articles) into an interactive knowledge graph where each document appears as a node, connected by relationship lines that reveal how they relate to each other.

With this visual network, you can:

  • Explore connections by clicking on any connecting line to see a summary of exactly how two works relate to each other
  • Dive deeper into specific papers by selecting them for detailed discussion with an AI research assistant
  • Discover hidden relationships through similarity scores that reveal which works share related topics, methods, or conclusions

CiteTrace makes complex research landscapes navigable at a glance, allowing you to trace how ideas evolve, methods connect, and research themes develop across your entire collection of materials.

How We Built It

  • Intel Tiber Compute

    • Developed and deployed a similarity-score model on Google Cloud Run
    • Tested and compiled multiple RAG systems

  • Frontend

    • Built with Expo Go, React Native, and TypeScript
    • Enabled smooth cross-platform deployment (iOS, Android, Web)

  • Database

    • Used Supabase for authentication and storage
    • Stored user-uploaded PDFs in Supabase buckets

  • Backend

    • Created a Python Flask REST API
    • Handled inference requests and file retrieval from Supabase

Challenges We Ran Into

  • Supabase
    We ran into persistent security issues accessing buckets, but after lots of trial and error, we found a way to bypass RLS and reliably upload and download PDFs

  • Graph Generation
    Using D3 to connect similar research papers often gave us messy, disconnected graphs. We spent hours refining the layout to keep nodes clustered and on-screen

  • PDF Parsing / Text Extraction
    Parsing and extracting meaningful content from diverse research PDFs was tricky due to formatting inconsistencies, but we built a pipeline to clean and structure the text for analysis reliably

Accomplishments That We're Proud Of

  • Clean, User-Friendly UI
    We are proud of the intuitive and user-friendly UI that allows users to upload PDFs easily and interact with the graph seamlessly

  • End-to-End RAG Pipeline Integration
    We successfully built and connected a full retrieval-augmented generation (RAG) system—allowing users to easily search for information they’ve seen across research papers

  • D3 Graph Optimization
    We overcame many layout issues—like drift, clutter, and inadequate color separation—to create an aesthetic, informative, and performant graph visualization

What We Learned

  • Supabase
    It was our first time using Supabase, so getting everything from bucket access to RLS configuration working was a fun but tricky learning experience

  • Hugging Face
    Hosting a Hugging Face model inside a real app was completely new to us, and integrating it smoothly took a lot of debugging and iteration

  • D3 Graphs
    Visualizing research paper connections with D3 was one of the hardest parts—making the graph both meaningful and readable took a ton of careful tuning

What's next for CiteTrace

  • Collaborative Workspaces
    Adding multi-user functionality will allow research teams to collectively build, annotate, and explore knowledge graphs, with real-time updates and permission controls

  • Expanded Document Support
    Beyond PDFs, we'll add support for web articles, slide decks, and even video transcripts, making CiteTrace a comprehensive research companion across all content types

  • Personalized Research Suggestions
    Building a recommendation system that analyzes your knowledge graph to suggest additional papers that would fill knowledge gaps or strengthen your understanding of specific topics

  • Academic Integration
    Creating plugins for reference managers like Zotero and Mendeley to seamlessly incorporate CiteTrace into existing research workflows and institutional repositories

Built With

Share this project:

Updates