HomSurfin was a startup that I co-founded in 2022. The idea was a real-estate mobile app with a focus on community and incorporation of large language models.

Co-Founder & CEO
Research, User Experience, Wireframing, Brand Design, App Development
Tech Used
VueJS, Quasar, Python, Postgres
Project Length
11 months

I met Caroline Swift Holden in mid-2022, when she was looking for a technical co-founder for her startup, HomSurfin. She'd already been working on HomSurfin part time for most of the year, and it was something she'd had in mind for years. We both had experience in the startup world already, and shared a common interest in the space.


The idea of HomSurfin would evolve over the next few months, but for the sake of simplicity, I'll describe the product in the sense of where we landed. In short, HomSurin was intended to have a multi-pronged approach to bringing modern tech to the world of real estate.

Social Aspects

One of the things missing from places like Zillow is social engagement. Users flock to places like Reddit, Instagram, and Pinterest to share homes and interior designs, when that is something that could be happening on the app itself. Currently none of the big apps feature the ability to post comments, create lists, write personal notes, etc.

Obviously there are some things to be taken into consideration here - like you don't want comments to just be people shitting on the house because that discourages anyone from buying. But realtors are busy, so making some questions public means they don't have to answer the same thing multiple times and it makes that information more available to others. For the phone-averse of us, it's also easier to write a question in a form rather than make a call.


There's also more that you can do here, though - with community engagement. You aren't just posting comments / questions for the realtor, but anyone in the community has the option to respond. You could promote neighborhood events, like a competition about who has the best Christmas decorations.

Another big area for folks is simply learning what different neighborhoods are like, because if you're moving into a new city, where do you go? Take the Boston area for instance: do you move to Allston-Brighton? Newton? Brookline? Cambridge? Somerville? Jamaica Plain? Roxbury? Chelsea? Malden? Medford? There are a lot of options and getting more insight into what the area itself is like helps narrow that down. While Zillow has things like the "walk score" and map, it lacks that personal touch of what the area is actually like.


Here is where we start getting into the LLM aspects. One of the early pitches for HomSurfin (long before I came onboard) was "Tinder for Houses". The idea was that it should be easy to signify interest or disinterest and take a mobile-first approach.

But there was more we could expand on there, too. When you express interest or disinterest in a property, the app could use that to start "learning" what you like and don't like. Recommendations could then be more tailored. The idea was that having something pitched to you that you love means less time searching means an opportunity to jump on the one you like much faster.

This can also tie in to the neighborhoods aspect. If you're liking a lot of homes in an area, hey, you might like that area. Conversely, Caroline had a friend who essentially wanted a log cabin. They didn't much care if it was in Maine, northern Vermont, Washington, even Alaska. When you're focused on something other than a specific geographic location, that also opens up a lot of opportunities to approach it differently. I want to live near the ocean, I'm not necessarily drawn to a specific city to achieve that.

Filling in the Blanks

Something that we discovered is exactly how fragmented real estate is. Places like Zillow / Trulia, Redfin, etc pull their data from MLS organizations. Simple explanation: MLS (Multiple Listing Service) is a database of homes. This contains information like the number of bedrooms, listing price, when it was last sold and for how much, square footage, etc.

What I hadn't known is that MLS organizations are often regional within a state - such that there are roughly 600 MLS orgs. They often use different software themselves and there's a very distinct lack of cohesion between them.

This is all to say that most of the time the listing agent is responsible for filling out information and sometimes they simply don't know or have that information. Something that people would love to search for is a house's architecture style - but you can't do that - places like Zillow don't even bother to expose it since it's so rarely entered into the MLS to be able to filter on.

One of the ideas here was to use the LLM to fill in some of those blanks: To have it be able to analyze the photos to tell whether a house is a tudor, or a colonial revival ranch, or a victorian, and actually expose that as something else that could be searched and filtered by.


Something that you see at times with listings is the ability to do a 3D walkthrough of the property. As it is today, this has a heavy cost associated with it because it has specific cameras that are required to generate that.

But what if your LLM could simply extrapolate from photos? Doorframes are (generally) a pretty standard size, so it could use that to measure walls, put them together based on similarity. Having a floorplan would make the process much easier, but perhaps it would be doable even without, making that an option more available to sellers of all tiers. Especially when you're looking at moving to a new city - the idea of getting a feel for the house more than photos is hugely beneficial.

Re-Imagining Rooms

This was an idea that I was particularly proud of. One of the things I have experienced myself when looking at photos of house interiors is an inability to imagine it in a different framing.

What is there can be gaudy, or kitschy, or just not what I would do. This idea was to use an LLM to be able to edit a photo of a room to show what it would look like in a different style. Change the furniture, wallpaper, drapes. Show me this room more minimalist. Okay more modern. More classic. More rustic. I would learn later that this aspect had already been built as interiorai .

HomSurfin screens

The Early Days

The first few months were a whirlwind. Caroline quit her day job to focus on this full time and I was doing it part time myself. We got into the BETA accelerator program in Minneapolis and things were moving quickly. It wasn't long before I hit a really big challenge: I had worked on building out startups before, but never quite so early. I was the 6th hire at Society of Grownups, but the others had been working full-time for months. Things like the branding, design, wireframing, prototyping, all of that was already done.

I realized quickly that there is a real balancing act where you want to move quickly but you also want to do things correctly. And especially where I was working on it part time, I was struggling with that. I pride myself on being good at multitasking, but there was just so much competing for attention.

Color themes on a pitch.

For instance: After working at a design agency for 4 years, I felt strongly about the idea of brand identity. This was something we never cracked at HomSurfin. We were in an awkward position where we wanted to appeal to younger buyers and older realtors both. Things like the name, logo, voice, tone. They were important to get right. So I wanted to work on that. I also wanted to make a styleguide. I also wanted to make progress with our prototype. I wanted the prototype to have unit tests. I wanted to figure out what hiring might look like. Everything I wanted to do competed against a thousand other things - all of them important.

Nondini Naqui was the founder of Society of Grownups and someone that I'd worked closely with and stayed good friends with and she was the first person I went to for help. I asked her, how do you balance those things? When everything feels extremely important? And she gave me perhaps the most useful advice I've ever gotten about anything.

This is very normal and frustrating. Just think about it this way: Importance vs urgency. Not everything important is urgent. Break out time to think strategically. Otherwise you could be busy but never add value.”
— Nondini Naqui

"Busy but never adding value" is the best way to describe how I felt about making my 30th color scheme in a given week. How Nondini phrased it helped me to recognize the importance of "doing the next important thing" (something I also talk about in my Context Switching blog post).

This advice would become a guiding star, not just in HomSurfin, but in other areas of my life, too. It's easy to feel overwhelmed and being able to quickly prioritize is a skill that I wish I had learned so much earlier.

An Alpha Test Event

December and January was a flurry of productivity. Caroline was doing fundraising campaigns and I was working on getting us an alpha of the app to be able to play with more realistically than the wireframing we'd been doing earlier in the autumn. And we began to talk about a live event.

In February I went out to Minneapolis for almost a week. I met Caroline in person, we did a bunch of work, we had our graduation from BETA, and the kind people at Twin Ignition were very accommodating to us hosting a private launch event.

The event was a way for announce our existence, do some networking, and more importantly, get people trying out the app. We'd put in enough houses (manually, since we didn't have MLS access yet) that you could easily search for places in the area, and we had our ~50 guests create accounts, search, like homes, leave comments, DM each other, and in general just try things out. We also had a room set aside for some workshopping / brainstorming for anyone who was interested in helping shape some new features - from the realtor dashboard to some of the community aspects we'd started wireframing but hadn't yet built.

The event was a huge success and a ton of fun. It gave us a lot of things to work on as well as the assurance that yeah, people liked this. A lot of what we wanted to do wasn't even present and it was still resonating quite well. We were so energized to continue building it out.

So what went wrong?

Ah, the age-old question. Both everything and also somehow nothing. It was a constant challenge and we made a lot of mistakes on the way, but none of that was what caused us to pull the plug.

In March and into April we were meeting with a lot of potential investors and going through interviews with more accelerator programs - TechStars, YCombinator, Gener8tor, others. As someone who is a shy introverted person, participating in those pitches is one of the scariest things I've ever done. We were also meeting with other founders in the real estate space.

Every founder we spoke to said the same thing:

I wish I had done literally anything else.”
— Every Real-Estate Startup Founder

They didn't mean they wished they weren't startup founders, but that they had gone into any other industry. The MLS organizations being fragmented was a problem. People who worked for them were distrusting after what they'd experienced with Zillow. It was really difficult not just to break into but to stay in.

That alone wouldn't have been enough to deter us, but the biggest thing we found in our investigations is that there isn't as much money in the space as you might think. For instance: The US has some 2 million real estate agents, but only about 75,000 of them pay for Zillow.

Revelations like this tanked our projections. The conclusion we came to was that we could still make a successful business. It would take years and we'd have a lot of trouble with funding but it would be doable. But it would never be what people think of as a Startup. Disruptive, sure, but without the kind of growth that is associated with startups.

With all of that in mind and the knowledge that even places like Zillow are struggling, we talked for weeks and finally decided to table the project. This area is still ripe for change - not only on the user side, but the systems that MLS organizations use are also fragmented and outdated. Nothing is standardized, nothing is coherent or cohesive. There are laws around real estate that may be changing that could have a big impact on possibilities in the next few years, but unfortunately the conclusion for this at the moment is that we can do more in a different sphere.

I had a lot of fun and learned way more than I would have thought - not just about the formation of a startup, but about myself and where my strengths area. Would I do it again? Absolutely, in a heartbeat.

I glossed over a bunch of areas for space, but if you are interested in learning more about this project, Caroline has a case study herself , and she also made a YouTube show . It goes for 26 episodes, starting when she began working on HomSurfin full time and going up to us closing shop. It's an honest peek behind the curtain of creating a startup, showing the stresses and heartbreaks, the joys and excitement, and so many of the little undefinable moments in between.