Synth Lab
Don't just read research—watch it think. AI agent powered architect that transforms the static text of multiple arXiv papers into "living" structural models. Synth Lab is a multimodal "Live Lab Notebook" that doesn't just explain research papers; it architecturally reconstructs them. As the agent analysis methodology of one or more arXiv papers, it simultaneously "draws" the logic in real-time using interleaved D3.js and Mermaid.js diagrams.
Demo
Synth Lab — live walkthrough
Screenshots
.png)
.png)
.png)
.png)
.png)
Problem Statement
The Problem
Academic research papers on arXiv are dense, static documents. Most AI summaries and outlines are just walls of text that lose the complex architecture of a research paper. Readers must manually trace through complex architectures, methodology flows, and hierarchical relationships buried in text and raw figures — a cognitive bottleneck that slows down understanding and knowledge synthesis, especially in fast-moving fields like ML & AI, Markets, and Drug Discovery.
There was no automated way to instantly convert a paper's structural logic into an interactive, navigable diagram — forcing researchers and engineers to spend hours building mental models that an AI could generate in seconds.
Solution
The Solution
Synth Lab is an application with an AI agent pipeline built on the Google Agent Development Kit that parses arXiv papers end-to-end and synthesises their architecture into interactive, hierarchical diagrams — turning static research into explorable technical synthesis, meta and comparative analysis.
- Search a topic and get the latest research papers pulled from arXiv api; the Gemini agent autonomously fetches and parses the full paper text.
- Agent extracts structural components, methodology steps, and relationships via LLM reasoning.
- Renders live, interactive hierarchical charts in React.js, d3.js, and mermaid.js.
- Deployed on Google Cloud Run via Docker for scalable, low-latency access from anywhere.
- Full TypeScript + Node.js back-end with type-safe API integration between the agent and front-end.
- The application operates in 4 tabs:
- Technical Synthesis — 1 or more research papers are synthesised into a Mermaid.js Flow Mode chart and a Final Chart, providing a master architectural diagram that consolidates the entire paper into a high-level system overview suitable for research planning and experiment design. Flow Mode lets users visually "unpack" complex research steps, with Sub-diagrams available to drill deeper into specific technical modules.
- Research Agent — Provides further textual analysis and subgraphs for deeper topic understanding.
- Deep Dive — A multimodal view that interleaves D3.js lexical Bubble Maps, providing a meta-analysis of the paper's weight by quantifying word density and thematic importance visually.
- Conceptual Dive — Interleaved charts rendered by an agent that breaks down concepts across multiple research papers, delivering a topic-by-topic comparative analysis.
Future
Future
Next milestones focus on scaling the agent workflow, improving synthesis quality, and broadening visualization output formats so researchers can move from reading to structured understanding even faster.
- Search for particular papers and uploading your own
- Frontier research paper comparative analysis to map overlaps and differences in the research field itself
- Higher-fidelity and faster extraction for syntax, equations, module dependencies, and evaluation pipelines.
- Deep dive includes visualizations of how topics interrelate with citation structure
- Export pathways for diagrams and structured summaries into shareable research artifacts.
- Even more ambitions? Agent provides future experimental design suggestions
Sponsors
Google Gemini