Centi

Centi came out of wanting one honest picture of my money instead of five apps that each showed a slice of it.

I built a dashboard that pulls account data together, cleans up the transactions, and turns them into something you can actually read: where money goes, how it trends, and what changed this month.

Raw transaction data is technically information, but it isn't insight. It's inconsistent, badly labeled, and spread across accounts that don't talk to each other.

The goal was less "another budgeting app" and more a clear reporting layer: connect the data once, then let the dashboard do the explaining.

Dashboards die from density. Centi shows a few numbers that matter up front, and keeps the deep detail one click away instead of one glance away.

Every chart answers a question in plain words — "what changed this month?" — instead of assuming the user wants to read graphs for fun.

The core of the system is the data model, not the charts. Transactions are normalized into a consistent structure first, which keeps categorization, summaries, and the API layer simple.

FastAPI serves the data, PostgreSQL stores it, and React renders the dashboard. Most of the real work happens in the cleanup pipeline before anything hits the screen.

The messy part was the data itself: the same merchant showing up five different ways, refunds, transfers pretending to be spending. Getting categorization trustworthy took more iterations than any UI decision.

Centi became a working dashboard and a strong example of the systems side of what I do: API integration, data modeling, and turning messy inputs into reporting people can trust.

Next would be smarter categorization that learns from corrections, and simple forecasting — showing where the month is heading, not just where it ended.