Project Info
Inspiration
Every time a water pipe breaks, a street floods, traffic stops, and millions of litres of clean water are lost before anyone even knows what happened. Cities have aging pipes everywhere, and most of them can only react. They fix pipes after they burst, not before. But here is the thing. Cities already record every break that has ever happened. That history sits in open data, unused. So we asked a simple question. What if you could use it to find the pipes most likely to fail next, and act before they do?
What it does
Blue Grid finds the water pipes most likely to break weeks in advance, and gives crews a clear action plan to fix them first. It shows a city's whole pipe network on an interactive map, with each pipe colored by how risky it is. Click any pipe and you get its risk score, the reasons behind that risk, an AI action plan for the next 72 hours, and a simple explanation of why it is risky. The goal is simple. Fix the right pipes first, and save the water that would have been lost.
How we built it
We used a real city's open data, both its full list of pipes and its complete history of past breaks, to train a model that scores every pipe by how likely it is to fail. The score looks at how old the pipe is, what it is made of, whether it has broken before, and how exposed it is to flooding. Then we checked it against what actually happened. 73% of the pipes that really broke show up in our highest risk group. It is not guessing. It is built on real results. The map uses Leaflet over real pipe shapes. The AI features, the action plans and the explanations, run on Claude. A Browserbase cloud agent pulls fresh break records straight from the city's live data portal, so the system stays up to date as new breaks get logged. The whole thing is live. The dashboard runs on Vercel and the AI backend runs on Render.
Challenges we ran into
Joining two messy datasets, the pipes and the breaks, and matching them up by location was harder than expected. Getting the AI features to work on the live site and not just on our laptop meant building a separate backend. And doing all of this solo, in one night, meant a lot of fast debugging.
Accomplishments we're proud of
It is real, it is live, and it is honest. We did not claim some magic prediction. We built a tool that finds at risk pipes in advance and proves it works against real break data. And it is a full product from start to finish. Real data goes in, and live AI action plans come out.
What we learned
The most useful data is often already public. Nobody is just connecting it. The hard part is not fancy AI. It is turning old city records into something a crew could actually act on tomorrow.
What's next
Bring it to other cities, since the method works anywhere with open pipe and break data. Add weather and freeze and thaw patterns, which are the biggest cause of breaks in cold places. And move from fixing pipes before they burst toward planning maintenance well ahead of time.
This project did not link a GitHub repository.
Analysis
No indexed repository for this project, so there are no commit stats to show.
Technology
- AnthropicUnchecked
- CSSUnchecked
- FlaskUnchecked
- HTMLUnchecked
- JavaScriptUnchecked
- PythonUnchecked
- RedisUnchecked
- VercelUnchecked
No repository was indexed for this project, so these Devpost claims have not been checked against code.
AI coding agents
No repository was indexed, so agent usage could not be checked.
Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.
Codebase size
No repository was indexed, so there is no codebase to measure.
This project did not link a GitHub repository, so there is nothing to diagram.
This project did not link a GitHub repository, so its feature claims have not been checked against code.
Export this project's context (description, README, evidence, key source files) to chat with an AI agent elsewhere.