Terrace
Years: 2024
Title: Founder, Designer
Project: ARES & Mirror
For the last 8 months, I’ve been researching, designing, and building in the “Home” space. I recently moved from New York to Atlanta. In the process of buying a house here, I learned how complicated the process is. For example, I got a home inspection report and I didn’t know how much it would cost to actually fix anything on there. There’s no standard formatting for a home inspection report, and there are hundreds of thousands of home inspectors across the United States. Everyone has their own kind of way of doing things — it’s kind of hectic.
When I was renovating my house, I relied on Notion to keep track of all of the projects, materials, and services I was using because I couldn’t easily find a product that would do that for me. Terrace’s dashboard centralizes all of the information you have across PDFs and even paper files. It helps by recapping everything you need to know about your home’s details — the foundation type, what kind of electrical system you have, HVAC, everything — and then couples that with large language model insights.
Looking back, I saw an opportunity to apply AI and large language models to the homeownership experience. That took me down a rabbit hole which brought me to what I’m building with Terrace today. Terrace starts out by analyzing your inspection report, but then later it takes in all this data and forecasts issues before they come up. The model is very similar to the way we approach health care today. It’s preventative versus reactionary.
I recently finished a design residency at IDEO CoLab Ventures and was able to build an early version of Terrace that's more of modern home affordability calculator and property analysis tool. We built it using OpenAI’s API and several websites with public property records.
When I was renovating my house, I relied on Notion to keep track of all of the projects, materials, and services I was using because I couldn’t easily find a product that would do that for me. Terrace’s dashboard centralizes all of the information you have across PDFs and even paper files. It helps by recapping everything you need to know about your home’s details — the foundation type, what kind of electrical system you have, HVAC, everything — and then couples that with large language model insights.
Looking back, I saw an opportunity to apply AI and large language models to the homeownership experience. That took me down a rabbit hole which brought me to what I’m building with Terrace today. Terrace starts out by analyzing your inspection report, but then later it takes in all this data and forecasts issues before they come up. The model is very similar to the way we approach health care today. It’s preventative versus reactionary.
I recently finished a design residency at IDEO CoLab Ventures and was able to build an early version of Terrace that's more of modern home affordability calculator and property analysis tool. We built it using OpenAI’s API and several websites with public property records.
We decided to build the first version of Terrace as a home affordability calculator and property analysis tool. The idea was that we would start with our version of a "Zestimate" and iterate our way to the virtual real estate agent experience. By starting with this, we were able to knock out the OpenAI API integration and create buyer profiles.
The latest homepage was built using Webflow, Figma, MidJourney, and a little bit of Javascript.
The original plan was to try build a virtual real estate experience with end to end onboarding. Turns out that was way more work than expected. During the grooming process, we decided to break it up into smaller parts and build components we could reuse.
Using OpenAI's API, we'd allow potential buyers to chat with a virtual agent, paste a link to a listing, and walkthrough the offer creation process, step by step, with a real agent on standby to help at any moment.
Part of the onboarding experience was to create a full buyer profile. This helped give us a sense as to where someone was in the process and help get them the right resources at the right time.
One constraint we had to work around was making sure buyers were serious. This meant collected pre-approval letters and storing data on the client.
We'd communicate with listing agents via email, using an agent to check updates and sync the app status. Once an offer was accepted, we'd walk buyers through a step by step closing guide, connecting them with all the relevant contractors to close on time.