Inspiration
Vacant offices and the need for more housing led us to a question: could empty space in a commercial building have a second life as a home?
Answering that question takes more than finding unused space. Owners need to understand the building, the requirements that apply, the work involved, and the potential costs. That information is scattered across property records, scans, imagery, and regulatory documents.
We built SecondStay to bring those pieces together and make the first conversation with an architect more informed.
What it does
SecondStay helps owners explore converting vacant commercial space in Los Angeles into housing.
It combines building records, 3D scans, and aerial imagery to show what evidence exists for published conversion requirements. Each requirement is categorized as Evidence found, Possible issue, Not enough data, or Needs professional review, all with supporting sources.
Users can explore buildings and available suites in 3D, review the work called for, and download next steps to share with an architect. They can also adjust cost, rent, and project assumptions to explore financial scenarios.
SecondStay provides a preliminary evidence review. Final determinations remain with permitting authorities and licensed professionals.
How we built it
We built SecondStay around three complementary data sources:
- CoStar supplies building characteristics, tenancy information, and available commercial space.
- Matterport supplies 3D scans and interior details where available.
- Vexcel supplies aerial imagery and information about the building’s surroundings.
Our Python backend connects these sources to a shared building profile, evaluates evidence against a requirements checklist, and produces a list of proposed work. We also use OpenStreetMap for street-frontage information and vision-model analysis of panoramas where structured scan details are missing.
The web interface uses Three.js for the building view and connects the evidence checklist to cost and income scenarios. Source references and uncertainty stay visible alongside the results.
Challenges we ran into
One challenge was connecting datasets that describe the same property at different scales. A building record, an individual suite, and a scanned floor are related, but their measurements are not interchangeable.
Missing information was another challenge. A scan may reveal stairs without establishing whether they meet a requirement. We needed to distinguish an observed feature from a conclusion that requires professional review.
Financial estimates presented a similar problem. Missing prices can make a partial total look deceptively complete. We made omissions explicit and withheld payback estimates until the necessary income assumptions were supplied.
Accomplishments that we're proud of
We connected property records, interior scans, and aerial imagery into a single workflow that moves from evidence to practical next steps.
We’re particularly proud of keeping uncertainty visible. Users can see what supports a finding, what remains unknown, and who needs to confirm it.
We also connected the physical building assessment to editable financial scenarios, helping users explore how different assumptions affect a potential conversion.
What we learned
We learned that combining datasets requires careful attention to what each source actually measures. More data is useful only when its scope and limitations are understood.
We also learned that uncertainty belongs in the interface. Showing missing evidence and incomplete pricing makes the results more useful for planning.
Most importantly, we learned to frame the product around the next decision: what should an owner investigate, gather, or discuss with an architect?
What's next for SecondStay
Our next priorities are expanding the sourced price book, improving data coverage, and validating the workflow with property owners and architects.
We also want to automate more of the evidence-gathering process and make it easier to update an assessment as new documents or professional findings become available.
