GraphSense

Our app in action

Inspiration

Data visualization shouldn't require a PhD in statistics or years of experience with complex tools. We were inspired by the frustration of analysts, researchers, and students who spend hours manually creating charts, often choosing suboptimal visualization types that obscure their data's true insights.

We envisioned an AI-powered assistant that could instantly understand any dataset and recommend the perfect visualizations – like having a data science expert at your fingertips.

What it does

GraphSense is an intelligent data visualization platform that transforms raw data into meaningful insights through AI-powered analysis.

Users simply upload a dataset (CSV, JSON, Excel, or TSV), and our sophisticated 3-agent AI pipeline automatically:

Analyzes the data structure, patterns, and statistical properties
Recommends the optimal chart types from 10+ visualization options with confidence scores
Validates each recommendation with detailed reasoning and appropriateness scoring

The platform features an advanced infinite canvas where users can drag, organize, and interact with their visualizations. A smart minimap provides instant navigation across large projects, while viewport-aware chart placement ensures new visualizations always appear exactly where you're looking.

How we built it

We architected GraphSense as a modern full-stack application with performance and user experience as core priorities:

Frontend: Built with Next.js 15 and React 19, featuring a custom infinite canvas engine with precise coordinate transformations, smooth zoom (0.1x–5x), and real-time minimap navigation. We implemented extensive performance optimizations including React.memo, data sampling for large datasets, and RAF-throttled updates.
AI Pipeline: Developed a sophisticated 3-agent system powered by Google Gemini 2.0 Flash:
1. Enhanced Data Profiler – Performs comprehensive statistical analysis and pattern recognition
2. Chart Recommender – Evaluates ALL 10 chart types with intelligent data mapping
3. Validation Agent – Provides quality assessment and recommendation refinement
Backend: Python FastAPI with robust error handling, pipeline orchestration, and real-time progress tracking. Integrated with Supabase for persistent storage and implemented comprehensive health monitoring.
Performance Engineering: Implemented intelligent data sampling, memoized rendering, and viewport-aware optimizations to handle datasets with 50,000+ points while maintaining smooth 60fps interactions.

Challenges we ran into

Coordinate System Complexity: Creating an infinite canvas with accurate minimap positioning required solving complex coordinate transformations (viewport translations, zoom scaling, Y-axis inversions).
AI Response Consistency: Reliable, structured responses from Gemini across different data types needed extensive prompt engineering and JSON parsing fallbacks.
Real-time Performance: Rendering large datasets demanded stratified sampling, React optimizations, and chart-specific point limits.
Canvas Positioning Precision: Ensuring charts appear exactly where expected required viewport-aware placement, anti-stacking algorithms, and fixing edge cases in coordinate calculations.

Accomplishments that we're proud of

Intelligent Automation: AI system that provides expert-level visualization recommendations with transparent reasoning.
Performance Excellence: Achieved smooth 60fps even with massive datasets.
User Experience Innovation: Built an intuitive infinite canvas with pixel-perfect positioning and smooth navigation.
Technical Sophistication: Production-ready system with error handling, real-time monitoring, and robust state management.
Comprehensive Coverage: Supports 10+ chart types with automatic field detection – no manual configuration required.

What we learned

AI Integration Complexity: Importance of robust error handling, response validation, and graceful degradation when AI services fail.
Performance Optimization: Scale-ready user experience requires intelligent data sampling and chart-specific strategies.
Coordinate Mathematics: Deepened understanding of 2D transformations, viewport management, and zoom precision.
User-Centric Design: Technical power means little without intuitive, thoughtful UX.