Synaps

Inspiration

The primary inspiration came from the universal frustration experienced by every student and researcher: the literature review bottleneck. We recognized that research time is wasted on manual, repetitive tasks—summarizing PDFs, cross-referencing concepts, and trying to mentally connect ideas scattered across dozens of documents.

We saw a clear opportunity to leverage modern Large Language Models (LLMs) with great summarization capabilities, large context windows and fast inference for instant knowledge synthesis. The goal was to build a tool that fundamentally shifts the researcher's role from information manager to knowledge creator

What it does

Synaps is an AI-powered platform that converts disorganized documents into a dynamic, interconnected Knowledge Graph, providing instant conceptual clarity and accelerated research synthesis.

Synaps eliminates the need for manual synthesis.

Live Concept Mapping: Users upload papers, and the system instantly generates an interactive graph using clear color coding: Blue Nodes represent the source Documents, and Gray Nodes represent the Shared Concepts extracted from the text.
Instant Synthesis: Clicking any Concept Node instantly retrieves a curated summary of that concept, synthesized from all linked documents. This summary directly addresses the concept's definition, technical approaches, and cross-document comparisons.
Overlap Detection: The visual graph structure immediately highlights conceptual overlaps, showing researchers which ideas are central or contentious across the entire literature set.

How we built it

Challenges we ran into

AWS Bedrock configuration was not trivial Choosing a good model was hard, large models were more accurate but really slow.

Accomplishments that we're proud of

We successfully demonstrated the core value proposition: We can visualize literature and texts in a generated graph, clicking a Concept Node and receiving a high-quality, cross-document summary without a noticeable loading delay. This instant feedback loop is the ultimate win for research productivity. This proves that there are great use cases for tools using the relational properties of text.
Engineered Value: Successfully orchestrating two specialized LLMs workflow stages within a single, coherent, event-driven workflow. This is an advanced use of AI that provides genuine engineered value beyond a simple API call.

What we learned

The hackathon provided invaluable lessons in technical prioritization and user empathy.

Prioritize User Flow: We learned that optimizing for the user experience (the Synthesis jump) was critical. The high-speed caching was a late but essential architectural decision to guarantee a fluid research flow.
Focus on the Gap: By concentrating solely on the literature review synthesis barrier, we avoided feature creep and built a tool that solved one massive problem exceptionally well.

What's next for Synaps

Support for adding documents in existing graph: A knowledge base in ever expanding, it only makes sense that we can keep adding stuff! We didn't have time to implement this feature, but it is the obvious next step.
Collaborative Mapping: Implement real-time multi-user collaboration to allow teams to build a collective knowledge graph, leveraging high-throughput event streaming platforms.
Relationship Deepening: Introduce the ability to track more complex semantic relationships between Nodes (e.g., "Concept A Refutes Concept B," or "Concept A Preceded Concept B").
Export Functionality: Develop tools to export the synthesized summaries and citations directly into standard academic formats (e.g., LaTeX, Word), further streamlining the thesis writing process.