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

Corides - Smart Ride sharing & Booking

Inspiration

The idea for CoRides.ai came during a frustrating rush hour in Islamabad. While sitting in traffic, I noticed something interesting. Hundreds of cars were on the road, but most of them had empty seats. At the same time, people were standing on the roadside trying to find rides that weren’t overpriced. That’s when I realized the real problem wasn’t the number of cars on the road. The problem was the lack of an intelligent system that could connect people already traveling in the same direction. I wanted to build something that felt less like a rigid app and more like a helpful neighbor who understands the roads and helps people share rides easily.

What it does

CoRides.ai is an AI powered carpooling platform designed to make city commuting simpler and more human. Instead of filling out complicated forms, users can simply tell the app where they want to go using text or voice. With Gemini Live Voice, users can speak naturally and the system understands their request instantly, converting the conversation into structured ride data.

The platform uses smart route matching that analyzes a driver’s full path rather than just the starting point and destination. This allows passengers to join rides from intermediate points along the route, making better use of empty seats in cars.

To build trust between users, the platform also keeps a transparent history of conversations. Every voice or text interaction related to a ride request is stored so that both drivers and passengers have a clear record of what was agreed upon.

How I built it

I built CoRides.ai as a solo developer while experimenting with what I consider an agent driven development approach. I worked mainly inside the Antigravity SDK, which helped automate much of the setup for my Firebase backend. This allowed me to focus on the core logic instead of spending time configuring infrastructure.

For the intelligence layer, I used Gemini through Firebase Genkit. Gemini acts as the bridge between human conversation and structured data by interpreting what the user says and converting it into ride requests, routes, and location details.

The application interface was built with Flutter. I focused on creating a clean, map centered design so users can easily see available routes and ride options without unnecessary complexity.

Challenges I ran into

One of the biggest challenges was connecting the AI’s understanding of human language with the real time behavior of the app. It is one thing for an AI model to understand a destination from a sentence, but another challenge to immediately reflect that information in the app interface.

When the AI recognizes a location or ride request, the app needs to update map markers, route lines, and ride information instantly. Integrating Gemini Live Voice added another layer of complexity because voice input happens in real time. I had to ensure that speech recognition, AI intent extraction, and UI updates worked smoothly together without breaking the user experience.

To handle this, I created a custom orchestration layer that connects the AI outputs directly to the app’s state so that a simple conversation or voice command can instantly become a real ride request.

Accomplishments that we're proud of

One of the things I’m most proud of is making the AI assistant feel natural to interact with. Instead of acting like a traditional form based application, the assistant can ask follow up questions and help complete ride details during the conversation.

Another accomplishment is the smart geospatial matching system that allows drivers to pick up passengers from intermediate points along their route, making better use of available car seats.

I’m also proud of building the entire system from idea to deployment as a solo developer, including a fully working backend on Firebase and an interactive mobile application.

What I learned

Building CoRides.ai showed me that AI is most powerful when it feels invisible. When the system works well, users don’t think about the technology behind it. They simply describe where they want to go and the system handles the rest.

I also learned how to convert messy human conversations and voice input into structured logic that a real world service can rely on. This process of translating natural language into reliable application behavior was one of the most valuable parts of the project.

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

This project is only the beginning. The next step is to introduce a social trust score for users so drivers and passengers can feel safer sharing rides with people in their community.

I also plan to improve the pricing intelligence so that ride costs can better reflect real time traffic conditions and route demand.

Another goal is expanding the Gemini Live voice assistant so users can manage their rides completely through conversation, including booking, updating, and canceling rides without needing to navigate through the app manually.

The long term vision is to make CoRides.ai a smarter, community driven transportation network that helps reduce traffic, lower commuting costs, and make better use of the cars already on the road