Intersection.ai

AI-powered urban safety platform that visualizes collision data, generates instant safety audits, and collaborates with voice agents to redesign intersections faster than traditional planning.

Comment

Inspiration

Montreal's road infrastructure is shit. City planning is slow and abstract, and we (the taxpayers) don't get to see what's really going on behind. Countless accidents happen every day because of dangerous intersections. Cities lack the proper tool to visualize what a safe city really looks like

What it does

Intersection.ai is a dynamic 3D mapping tool that transforms hazardous intersections into safer spaces on demand. By selecting high-risk areas, users receive instant AI-powered redesign proposals featuring:

  • Enhanced signage and road markings
  • Separated cycling infrastructure and improved pedestrian pathways
  • Practical safety assessments with specific implementation steps
  • Caching system replaying the latest crash happened at EVERY specific intersection

Intersection.ai bridges the gap between raw traffic statistics and tangible solutions, enabling municipalities to identify priorities and execute safety improvements with greater speed and impact.

How we built it

Frontend Framework & UI

  • Next.js 16
  • TypeScript
  • Tailwind CSS 4

Mapping & Visualization

  • Mapbox GL
  • React Map GL

AI & Machine Learning

  • Google Gemini 2.5 Flash
  • Google Gemini Nano Banana (intersection redesign image generation)
  • ElevenLabs (real-time voice synthesis for persona agents)
  • Replicate API (AI video generation)
  • Gumloop API (Real-time web search data)

Backend & Data

  • Next.js API Routes (serverless endpoints)
  • MongoDB (collision data storage, cluster persistence and video caching system)

Challenges we ran into

  • Merging real-world collision data with AI-generated visuals while maintaining accuracy and clarity
  • Crafting multi-persona perspectives that delivered meaningful insights without overwhelming each stakeholder group
  • Maintaining smooth real-time visual updates and performance optimization within a web-based platform

  • Integrating video replay crash simulations that accurately depicted incident scenarios while remaining technically feasible and visually coherent

    Accomplishments that we're proud of

  • Developed a visually-driven urban safety platform that converts complex data into clear, actionable guidance

  • Delivered AI-generated intersection redesigns that balance realism, clarity, and contextual relevance

  • Established a multi-perspective analysis system that allows planners to understand safety challenges through the lens of cyclists, engineers, and policymakers

  • Created comprehensive safety audits that extend beyond visualization to offer specific, implementable recommendations for on-the-ground change

    What we learned

  • How to synthesize AI-generated imagery with geospatial and traffic datasets to build practical urban planning solutions

  • The critical role visual communication plays in driving safety-oriented decision-making

  • The art of striking a balance between real-world precision and AI-powered visualizations to create tools that are both engaging and credible

    What's next for Intersection.ai

  • Expand the map to cities across the entirety of Canada, and eventually Worldwide.

  • Preventive measures to predict where the next crash will happen

  • Collaborate with city governments and advocacy groups to implement Intersection.ai as a practical decision-support platform for tangible urban safety enhancements

Built With

Share this project:

Updates