Inspiration
Every single day, thousands of on-demand delivery couriers across Latin America juggle multiple apps (Uber Eats, Rappi, and DiDi Food) simultaneously. Their reality on the streets is chaotic: they ride with phones mounted to motorcycle handlebars, risking collisions while manually switching between apps, constantly guessing which order is worth their time, and frequently heading back empty-handed.
The courier's challenge is simple: how to maximize earnings for the courier.
We asked ourselves: What if couriers had an AI agent that orchestrated incoming orders across every platform in real time, batching routes without hurting service metrics?.
That was the spark behind VYGO: speed, connectivity, and real savings.
What it does
VYGO is a real-time AI agent and route orchestrator built for gig economy delivery couriers who multiapp across Uber Eats, Rappi, and DiDi Food. Instead of forcing couriers to manually switch apps on their handlebars or guess which deliveries are worth the trip, VYGO combines multiple active and incoming orders into a unified, high-yield itinerary without violating the Service Level Agreements (SLAs) or thermal food freshness constraints of any platform
How we built it
We built VYGO from the ground up to ensure low latency, screen clarity, and driver safety on the road:
- Mobile App: We created a lightweight web application that runs smoothly on mobile devices, featuring live maps to guide the courier step-by-step.
- Central Server: A backend server receives incoming trip offers, runs safety checks, and returns the best driving path in under 150 milliseconds.
- AI Routing Agent: We engineered an automated decision engine that evaluates physical city streets and traffic patterns. First, it runs basic feasibility checks to ensure meals fit inside the delivery backpack and will arrive fresh.
- Database & Location: We connected a cloud database to store driver accounts securely and record live locations every few seconds while on the move.
Challenges we ran into
- The Inverted Map Coordinates Bug: During early testing, our delivery stops in Monterrey mistakenly appeared in the middle of the Pacific Ocean. We discovered that our mapping tools and database read latitude and longitude in reverse order.
- Beating the 30-Second Countdown: Most delivery platforms withdraw an offer if the driver does not respond within roughly 30 seconds. In our earliest versions, calculating complex multi-stop routes took over a second, which was too slow.
- Scrapping the Chat Assistant: We initially considered building a conversational voice or text assistant for couriers. However, hands-on tests showed that talking or texting while steering a motorcycle in traffic is dangerously distracting and takes too long to respond.
- Handling Dropped Cell Signals: Mobile data connections frequently cut out on city streets. To keep the screen from freezing while riding, we added an offline safety net: if the server doesn't respond in time, the app falls back to simple built-in rules stored on the phone so the driver can keep moving without interruption.
Accomplishments that we're proud of
- 3.5 Times Higher Earnings: A courier using VYGO can earn $889 MXN in a two-hour shift compared to $254 MXN working the traditional way.
- 3.2 Times More Completed Deliveries: The driver completed 16 orders in two hours by bundling nearby deliveries, compared to only 5 orders done one by one.
- 54% Fewer Kilometers per Delivery: While the total shift distance increased slightly because the courier completed far more deliveries, the distance driven per order dropped from 9.10 km down to 4.17 km. This cut estimated fuel costs and vehicle wear by more than half for every single completed delivery.
What we learned
- Time, Not Orders, Is the Real Bottleneck: Couriers running multiple apps rarely suffer from a lack of delivery requests—they have too many. Accepting a mediocre delivery early can lock up their next half-hour, preventing them from taking much better orders that appear just minutes later.
- Smart Bundling Beats Driving Faster: Delivering one order at a time turns a workday into an endless queue of empty trips back to restaurant areas. Grouping orders that head toward the same neighborhood allows a single trip to generate multiple payouts.
What's next for Vygo
- Floating Phone Overlay: Turning VYGO into a native floating banner that stays on top of any delivery app, allowing drivers to accept suggested orders with a single tap.
- Live Traffic and Extreme Weather Alerts: Adding live updates for heavy rain and major road closures along primary avenues so VYGO can re-route deliveries before drivers get stuck in gridlock.
Built With
- ai
- dp
- fastapi
- maplibre
- maptiler
- ml
- postgresql
- python
- react
- supabase
- tailwind
- typescript
- vite
- zustand
Log in or sign up for Devpost to join the conversation.