PROPTECH / AI / B2B2C
Role:
Time Line:
2025
Responsibilities
Visual design, Research, Product Strategy,
OneHome is the home search portal used by MLSs across North America, with over 20 million monthly visits. Buyers have very specific needs, like a home without stairs, that filters can't capture, so they scrolled through every listing photo. I led the design of OneHome's first AI search experience, launched as an MVP to learn fast from real buyers.
AI trends: text input has the lowest barrier to entry; image search is familiar from Pinterest and Google.
Analytics: buyers spent a large share of their time in the photo gallery.
Survey: about 40% asked for more filters, even though we already had many.
"No basements. Must have dishwasher and bathtub."
More filters would make search more complex and still miss the next personal need. People needed to search in their own words and through what they see.
At the time, LLMs were still new and hard to integrate, while image similarity search was mature and proven. I designed two ways for buyers to search by what matters to them.
Solution A
Text search
Buyers describe what they want, like "no stairs, open kitchen."
Strength: the most natural way to search, with the fewest steps.
Trade-off: hard to integrate; our AI vendor couldn't support it yet.
Solution B
Image search
Buyers upload a photo of a home they love and get similar listings.
Strength: reliable technology that matches how buyers decide, through photos.
Trade-off: buyers need a reference photo to start.
For technical reasons, the team chose image search as our MVP. It let us bring AI to buyers quickly, get early feedback, and learn which criteria matter most, insights we could use to improve our filters. I designed it to be accessible, like detecting stairs in photos, and resilient, with guidance when an image is too large or no homes match. Text search was documented as the next step, ready for when the technology matured.
Strong MLS adoption and engagement after launch.
Early feedback that shaped the roadmap: the MVP showed that everyday criteria are easier to type than to show, making text search the clear next step.
A foundation for AI in OneHome: more AI features followed, like AI redesign of gallery photos, and the team is exploring new AI providers for text search.