Did AI forget homebuilding?
Why static CAD drawings aren’t data? And what it actually takes to build spatial AI that’s construction-ready by design.
AI has rewritten how we interact with words, code, and data, but it hasn’t touched homebuilding. Frontier models can write blog posts and legal briefs, but they can’t reliably reason about a floor plan, a load-bearing wall, or what happens when you move a wall two feet. That’s not a gap more compute will fix; it’s a data problem. Homebuilding’s most valuable knowledge- how a lot width constrains a garage, how a change on one drawing should ripple everywhere it matters- has never been written down anywhere a model can reach it.
You’ll see how homes represented as living, structured data instead of static drawings lets builders automate estimating, sales configuration, and field documentation from a single source of truth.
Meet the speakers



Vaidas Razgaitis is a Senior Research Engineer at Higharc, where he works on computer vision, LLMs and multi-agent systems with spatial intelligence. He holds a master’s degree in structural engineering from Columbia University and worked on skyscrapers and airport megaprojects in Mexico City, Singapore and New York City. He also taught high-rise design as an adjunct professor at Pratt Institute’s School of Architecture. Later, he studied computer science at the University of Chicago, which pulled him into machine learning and generative AI. He is a co-author of Tokenizing Buildings: A Transformer for Layout Synthesis, a 2025 paper on transformer-based layout synthesis for buildings.
Watch On-Demand

.png)
.png)

.png)

%2520(780%2520x%2520540%2520px).png)




%2520(4).png)
%2520(1).png)
.png)