Corides.ai - A Carpooling service powered by Gemini Ai

Corides.ai - A Carpooling service powered by Gemini Ai
Corides.ai - A Carpooling service powered by Gemini Ai

Inspiration

It started during a particularly frustrating rush hour in Islamabad. Looking around, I saw hundreds of cars, most with three empty seats, while people stood on the curbside hoping for a ride that wasn't overpriced. I realized the problem wasn't a lack of cars, but it was a lack of a "brain" to connect them. I wanted to build a community-driven service where technology felt less like a rigid app and more like a helpful neighbor who knows the roads.

What it does

CoRides.ai is an AI-powered carpooling ecosystem I designed to make city travel more human. Conversational Booking: Instead of fumbling with complicated forms, you just tell the app where you’re going. Smart Route Matching: I didn't just match start and end points. My system looks at the driver’s entire path, including intermediate waypoints, to find the perfect fit for a passenger. Transparent History: Every voice and text interaction is saved, creating a clear and trustworthy record of ride agreements.

How we built it

To pull this off as a solo developer, I treated this project as an experiment in the future of "Agentic Development." The Tools: I lived inside the Antigravity SDK, using its agentic features to handle the heavy lifting of setting up my Firebase backend. This allowed me to focus on the core logic while the IDE managed the infrastructure. The Intelligence: I used Gemini 3 via Firebase Genkit. It acts as the "translator" between what a human says and the structured data my database needs. The UI: Built with Flutter, I focused on a clean, map-centric experience where information is always easy to find.

Challenges I ran into

My biggest "lightbulb" moment and my toughest challenge was building the bridge between the AI’s text-based brain and my Flutter widgets. It’s one thing for Gemini to understand a destination; it’s another to have that data instantly update a map marker, change a price tag, or trigger a UI transition in real-time. I had to create a custom orchestration layer to ensure that the "intent" extracted by the AI was flawlessly mapped to the app's state, turning a conversation into an actual interactive experience.

Accomplishments that I'm proud of

Making AI Feel Natural: I am incredibly proud of how the Gemini assistant feels it doesn't just process data; it asks the right questions to fill in the gaps. Smart Geospatial Matching: Successfully implementing the logic that allows drivers to pick up passengers from intermediate waypoints, maximizing car utility. Solo Efficiency: I'm proud of moving from a blank screen to a fully deployed Firebase backend by effectively utilizing Antigravity’s automation tools.

What I learned

Building CoRides.ai taught me that AI is most powerful when it’s invisible. I learned how to manage the "fuzzy" nature of human speech and turn it into the hard logic required for a ride-sharing service. Most importantly, I learned that with the right AI tools, a solo developer can build complex, full-stack architectures that previously required entire teams.

What's next for Corides.ai - A Carpooling service powered by Gemini Ai

I'm just getting started. My next steps involve adding a "Social Trust" score to user profiles and refining my pricing AI to better reflect real-time traffic conditions in cities. I want CoRides.ai to be the reason there's one less car stuck in traffic tomorrow.