Palmetto wants software developers to electrify America using its AI building models

For someone who wants solar panels, the question is often: “How many?” And for a heat pump, it’s “how big?” Answering those questions typically requires a contractor to set foot on your property. 

Yet over the last decade, solar installers have trimmed that time consuming process considerably. Some, like Tesla, have eliminated it entirely. The secret is software — which, when coupled with a range of data sources, allows installers to model exactly how many panels a house will need. 

Like many other solar installers, Palmetto has its own internal tool to assess potential projects, which is built on software it acquired when it bought Mapdwell, a solar mapping startup.

But as growth has slowed in the market for residential solar, the company has begun to tackle other electrification projects, including backup batteries and heat pumps. Developing software for those products, which live inside the home where satellite imagery can’t reach, is trickier.

So Palmetto started gathering data to develop a virtual view of every residential building in the United States. Palmetto’s president of energy intelligence Michael Bratsafolis told TechCrunch that the company is “basically simulating digital twins of the entire US residential building stock.”

The company’s tool relies on public and private data. Where it doesn’t have exact information, its AI uses clues to infer what might be lurking in the walls. For example, a house built in the 1950s likely used 2×4 lumber for the exterior walls, limiting the amount of insulation that can fit inside.

“This technology can break the home down into like 60 plus different characteristics and attributes,” Bratsafolis said.

To validate the models, Palmetto can turn to the data it has already gathered on homes that have bought or are leasing solar panels from the company. 

“That allows us to anonymize and use the real information and data about a home to compare how the model is performing and help train the model further,” he said.

Palmetto has also decided to open the tool to outside developers, a move that Bratsafolis said was inspired by his time at Twilio.

“I came from that developer motion, the micro-service approach to providing API building blocks to kind of empower partners and developers,” he said. 

Developers get 500 calls free every month, and after that it’s five cents per call. For bigger customers, Bratsafolis said the company is open to negotiating special pricing.

Bratsifolis said he hopes developers will build tools that speed electrification of US building stock, and Palmetto will get some early clues about where the market is headed: “You just don’t necessarily know what the use cases could be. And with an API product, you get incredible insight into the demand and the market.”