DepotOS

Inspiration

We ride Waymos often and kept wondering what happens between rides. Who charges the car? Who cleans up after the last rider? What happens when someone leaves a bag or spills a drink? When we looked into it, we found that robotaxis drive themselves, but the depots that charge and clean them still run almost entirely on people. That's why we came up with DepotOS.

What it does

DepotOS is an operating system for robotaxi depots. It decides, car by car, what a robot can handle and when a person is actually needed.

1. Detect: cameras and odor sensors check every car as it arrives.
2. Assign: charging and simple cleaning go to robots. Bags, big messes and damage go to a person.
3. Decide: if the fleet is covered for the morning, non-urgent jobs wait for the morning crew. If not, a worker is called in.
4. Protect: hazard cars are isolated, and workers get a briefing before opening a door.
5. Forecast: it predicts busy nights, like Saturdays, so the right crew is booked ahead.

How we built it

We pieced together how depots work from published reporting, competitor launches and forecasts, then built a simple cost model for a 50-car depot. We built the prototype, an interactive control-room web app, using Replit and Claude Code. All data in the demo is simulated.

Challenges we ran into

The biggest challenge was limited data. Companies like Waymo don't publish staffing, schedules or how often cars need cleaning, so we had to make assumptions and label them clearly as estimates. Some of the biggest numbers came from competitors, so we had to double-check what we could trust. Building the app with Replit also took a lot of back-and-forth to get both the design and the logic right.

Accomplishments that we're proud of

Building a web app from scratch for the first time, going deep into an industry we knew little about, and taking an idea all the way to a working prototype and a business model.

What we learned

Planning precisely before building saves a lot of time, especially with AI tools. We also learned how to turn messy information into something easy to understand, and what goes into thinking through a startup, from customer to pricing.

What's next for DepotOS

Interview depot workers and fleet managers, test our pricing with a few partners, and run a 60-day pilot at one LA depot where DepotOS recommends and people decide.

Built With

• chatgpt
• claude
• powerpoint
• replit