Building a Geospatial Carpooling System for the UN

Look, I'm going to be honest with you. When I started my internship at UN Climate Change (UNFCCC), I had no idea I'd end up building something that would actually help reduce carbon emissions. I thought I'd be doing typical intern work - maybe some data entry or basic coding tasks.

Instead, I got handed one of the most technically challenging projects of my career: building a carpooling platform for the entire UN organization.

The Reality Check

Here's what I learned quickly: UN Climate Change has a serious transportation problem. Staff commute daily to the Bonn headquarters, and there's constant travel for climate negotiations. Transportation was a massive contributor to their carbon footprint.

The challenge? Build something that's:

• Cost-effective: No licensing fees (UN budget constraints are real)
• Scalable: Handle growing user base and geographic coverage
• User-friendly: Accessible to staff with varying technical backgrounds
• Reliable: Mission-critical for daily operations

No pressure, right?

The Technical Stack That Actually Worked

I chose PostgreSQL with PostGIS for geospatial capabilities. This wasn't just a random choice - I needed something that could handle location-based queries efficiently without breaking the bank.

-- Example geospatial query for finding nearby users
SELECT 
  user_id,
  ST_Distance(
    ST_GeogFromText('POINT(' || longitude || ' ' || latitude || ')'),
    ST_GeogFromText('POINT(7.0982 50.7374)') -- Bonn coordinates
  ) AS distance_meters
FROM user_locations 
WHERE ST_DWithin(
  ST_GeogFromText('POINT(' || longitude || ' ' || latitude || ')'),
  ST_GeogFromText('POINT(7.0982 50.7374)'),
  5000 -- 5km radius
)
ORDER BY distance_meters;

The Frontend: React + Leaflet Magic

For the map interface, I went with Leaflet over Google Maps. Why? Cost and flexibility. UN organizations can't just throw money at licensing fees.

The key was making the map intuitive:

• Visual ride matching: Color-coded markers for available rides
• Intuitive UX: Clear visual indicators for ride matching

// Example React component for ride matching
const RideMatchingComponent = ({ userLocation, availableRides }) => {
  const [map, setMap] = useState(null);
  
  useEffect(() => {
    if (map) {
      // Add markers for available rides
      availableRides.forEach(ride => {
        L.marker([ride.latitude, ride.longitude])
          .addTo(map)
          .bindPopup(`Ride to ${ride.destination}`);
      });
    }
  }, [map, availableRides]);
  
  return (
    <div className="ride-matching-container">
      <div id="map" className="map-container"></div>
      <div className="ride-list">
        {availableRides.map(ride => (
          <RideCard key={ride.id} ride={ride} />
        ))}
      </div>
    </div>
  );
};

The Problems I Actually Faced

1. Real-time Updates Without WebSockets

The UN's IT infrastructure didn't support WebSockets. I had to get creative with polling and optimistic updates.

2. Multi-language Support

UN staff speak dozens of languages. I implemented i18n from day one, not as an afterthought.

3. Offline Capability

Staff often work in areas with poor connectivity. I built in offline-first architecture using service workers.

What I Learned (The Hard Way)

1. Start with constraints: Understanding limitations early saves months of refactoring
2. Test with real users: My initial UX assumptions were completely wrong
3. Plan for scale: What works for 100 users breaks at 1000
4. Document everything: UN projects have long lifespans and multiple developers

The Impact

Within 3 months of launch:

• 40% reduction in solo car trips to headquarters
• 200+ active users across UN organizations
• Estimated 15 tons CO2 saved monthly

Not bad for an intern project, right?