RideMates

Inspiration

Inspiration I commuted to De Anza by bus for months, watching classmates drive past me alone in their cars, getting there in a fraction of the time while constantly complaining about gas prices, with three empty seats to spare. De Anza even has a carpool program with reserved parking, but matching is entirely manual, so barely anyone uses it. RideMates fixes that mismatch: faster, cheaper rides for riders, split gas costs for drivers, fewer cars on the road, and real friendships between students who were going the same way all along.

What it does

RideMates matches De Anza students with classmates on the same route and schedule using DigitalOcean's Gradient AI not just nearest-match, but genuine route and time overlap, explained in plain language. Each match shows an estimated carbon savings. If no carpool match exists, RideMates suggests a free VTA transit route instead.

How we built it

Next.js, TypeScript, and Tailwind on the frontend; a Node backend with DigitalOcean Managed Postgres for riders, requests, and matches; DigitalOcean Gradient AI for the matching logic and explanations; Leaflet + OpenStreetMap for maps; deployed on DigitalOcean App Platform.

Challenges we ran into

Prompt engineering for the matching logic took real iteration getting Gradient AI to reason about genuine route/time overlap instead of just returning a plausible-sounding but arbitrary match. Coming from AWS, the DigitalOcean dashboard had a learning curve of its own different mental model, different naming, figuring out where things lived. We also went through a few rounds trying to find an auth provider that actually fit cleanly into the stack without adding a ton of overhead.

Accomplishments that we're proud of

Getting to solve a problem I actually lived — the bus commute, watching classmates drive past alone — instead of building something abstract. Beyond that, we're proud of getting real AI-reasoned matching working on Gradient instead of a naive nearest-neighbor algorithm, shipping the VTA fallback so the app stays useful even without a carpool match, and getting the whole thing running end-to-end on a new-to-us stack (DigitalOcean) within the weekend.

What we learned

Navigating a new cloud ecosystem (DigitalOcean) after AWS, prompting an LLM for real reasoning instead of just text generation, scoping a big idea down to something buildable, and that auth is never as simple as it looks.

What's next for RideMates

Multi-rider pooling, a recurring "commute buddy" feature, a pilot with De Anza's carpool program or DASG, and expansion to other 100%-commuter California community colleges.

Built With

• ai
• digitalocean
• doctl
• gradient
• next
• postgresql
• react
• tailwind