Predictive intelligence for every listing.
Real estate agents have been pricing on gut feeling, and it's costing them client trust. Stairtop is the MLS-trained platform I built from scratch — brand, product, and design system — to give agents the evidence to back up what they already know.
Agents were pricing on gut feeling.
Real estate agents make some of the highest-stakes recommendations in a client's life — what to list at, when to move — largely on instinct and experience. That's a hard thing to defend when a client pushes back, and it's a harder thing to scale. Client trust erodes fast when "trust me" is the only evidence on the table.
Stairtop's premise: give agents predictive, MLS-trained intelligence for any listing — not to replace their judgment, but to back it up with market evidence the client can see for themselves.
I built the brand and the product.
As the sole designer, I owned everything: the name, the logo, the color system, the entire product experience, and the design system underneath it. Every prototype that got tested — and every pixel that shipped — came out of my own process, from the first sketch of the icon mark to the final production UI.
They didn't want more data. They wanted one number.
My earliest prototypes tried to show agents everything at once — every pricing scenario, every band of confidence. In testing, we learned fast that what agents actually leaned on, over and over, was simpler: the days-on-market curve. How long will this sit if we price it here versus there? That's the conversation they have with clients, so that's the number we leaned into — sharpening the pricing simulator until that one curve carried the whole story.
The evidence agents can actually use.
A quick walkthrough of the product.
Built from scratch. Now scaling.
There was no existing brand or product to build on — every part of Stairtop, down to the coffee we hand out at events, came out of the same design system. That consistency is part of why agents trust it: it feels like one considered thing, not a patchwork of features.
We're now expanding into new states, taking the same predictive model and brand into markets beyond where we started.
Simplicity was the hardest thing to design.
It would have been easy to keep adding bands, charts, and confidence intervals because the model could support them. The real work was figuring out which single number actually changed a conversation with a client — and having the discipline to build the whole product around that number instead of around what the data could technically show.
