Urban Nexus

Business Mapper

Comment

Inspiration

Finding the “right spot” is often the make-or-break moment for cafés, pharmacies, and other SMBs. While researching Indian startup failure rates, we discovered that up to 70% of closures trace back to poor location choice. As students in Computer Engineering and BS Data Science, we saw an opportunity to apply geospatial data, machine-learning, and sustainability metrics to a real-world pain point. Urban Nexus Business Mapper (UNBM) was born from the desire to give every entrepreneur “big-consultancy” insight—without the big-consultancy price.

What it does

UNBM is a web-based SaaS platform that:

  • Ingests Google Maps, census, traffic, and open-GIS layers.
  • Runs an AI scoring model that weights foot-traffic, competition density, logistics cost, and even carbon impact.
  • Presents a ranked shortlist of top-3 optimal locations, complete with heatmaps, cost estimates, and sustainability scores.
  • Generates a shareable PDF so founders can pitch landlords or investors with data-backed confidence.

How we built it

  • Frontend: React.js + Tailwind on Vercel for rapid UI iterations.
  • Backend: Node.js micro-services on Firebase Functions; REST & WebSocket endpoints.
  • Data stack: Google Maps API for POIs and routing; OpenStreetMap tiles; Gemini LLM for narrative insight; Pandas & GeoPandas for on-the-fly feature engineering.
  • Model: Gradient-boosted trees that output a composite score
    [ Score = 0.35\,T_{\text{foot}} + 0.25\,C_{\text{demand}} + 0.2\,L_{\text{logistics}} + 0.2\,S_{\text{carbon}} ]
  • DevOps: GitHub Actions → Firebase Hosting → Cloudflare CDN for painless CI/CD.

Challenges we ran into

  • Data sparsity: Detailed retail foot-traffic datasets for Pune were scarce. We improvised by proxying Google Popular Times and weekday GPS pings.
  • API cost ceilings: Free-tier call limits forced us to batch requests, cache results, and implement a spline-based traffic estimator to avoid overages.
  • Balancing accuracy vs. speed: Early models took 3min per location. Pruning irrelevant features and moving heavy GIS joins to PostGIS cut this to 11s.
  • Explaining AI to non-tech users: We added tooltip explanations and a “Why this score?” accordion to build trust.

Accomplishments that we're proud of

  • Deployed a working MVP in eight weeks that demoed live at Innothon and landed in the top-5 finalists among 120 teams.
  • Integrated a sustainability module that estimates (\Delta \text{CO}_2) emissions if logistics routes shift—turning mere site selection into green site selection.
  • Achieved 92% agreement with human experts on a validation set of 40 existing retail outlets.

What we learned

  • Real-world datasets are messy; strong ETL pipelines matter as much as fancy algorithms.
  • UI/UX trumps raw tech: early testers valued clear explanations over marginal gains in model precision.
  • Cross-disciplinary thinking—combining urban planning, GIS, and machine learning—creates solutions that single-domain approaches miss.

What's next for Urban Nexus

  • Mobile Companion App: On-site AR overlay to visualize score heatmaps through a phone camera.
  • City Roll-out: Add Mumbai and Bengaluru using scalable ETL templates.
  • Dynamic Rent Index: Partner with listing platforms for real-time rental feeds, refitting our cost component.
  • API Monetization: Expose a /score endpoint for prop-tech and food-delivery startups.
  • Research Collaboration: Work with municipal bodies to incorporate UNBM's carbon-aware routing into urban planning dashboards. Urban Nexus is our step toward data-democratized city building—where every entrepreneur can make location decisions that are both profitable **and* sustainable.*

Built With

Share this project:

Updates