AI Infrastructure Dashboard

What it is

A dashboard that puts AI infrastructure — capacity, spend, and utilization — into a single view. The goal was legibility: someone should be able to open it and know within a few seconds whether anything needs attention.

I built it end to end on my own, which meant making every decision, including the ones I'd have preferred someone else make for me.

Why I built it

The underlying data for this kind of thing is usually spread across several places and none of them agree on what a "unit" is. I wanted to see whether I could turn that into one screen that was actually useful rather than one screen that was merely dense.

What I learned

• Most of the work was deciding what to leave out. The first version showed everything I had. It was unreadable. The useful version shows less.
• A dashboard is a design problem before it's an engineering problem. Choosing the comparison — versus last period, versus budget, versus capacity — determines whether a number means anything. I got that wrong before I got it right.
• Shipping it myself meant owning the boring parts: deploys, error states, what the page looks like before data loads. Those took longer than the charts did.

Where it stands

Shipped and working.