Hadla Bergman Is Making Building-Design AI Work Across Different Engineering Rulebooks

A seventy-story high-rise in the United States exposed exactly the kind of problem Hadla Bergman is responsible for solving. The engineering firm approached lighting design differently from the convention Endra's platform had already encoded. Both approaches were reasonable, but they could not simply coexist without the software understanding how and when each should apply. Bergman, a Founding Engineer at Endra, had to determine whether the difference was a narrow exception or evidence that the system needed a more general rule capable of handling other projects it had not yet encountered.

"That is the interesting part of this work," Bergman said. "The question is rarely which convention is right. The real question is whether the system can represent the reasoning behind both and know when each one applies." On a large project with tens of thousands of elements, a local workaround may solve one case while creating another problem elsewhere in the model.

Endra recently raised a $50 million Series A led by Andreessen Horowitz (a16z), following its recent $20 million Seed round and bringing its total funding to $74 million as the company expands internationally. Bergman leads the technical integration behind Endra's expansion into the United States. The company builds AI and software for mechanical, electrical, and plumbing design, often referred to as MEP, covering systems such as power distribution, lighting, fire systems, and containment. Her work involves adapting the platform to American codes, standards, and design workflows while preserving support for engineering practices already used in Sweden and the United Kingdom.

A concrete example is Bergman's work on cable-conduit modeling. While working through a complex engineering problem, she developed a novel algorithm for handling complex cable-conduit bends.

"What excites me isn't just solving the problem in front of me, but pushing the boundary of what the product itself can do."

For Bergman, that kind of problem-solving is almost reflexive. She has a tendency to keep pulling at a difficult engineering problem until she can see the computational structure underneath it. She describes it as "thinking in code": once she understands the problem deeply enough, she starts seeing the algorithm that could make a previously difficult problem solvable.

The cable-conduit problem was one such case: she reduced a highly complex physical modeling problem to a computational approach that could be expressed systematically in the product. That ability to find structure inside complexity is one of the qualities that distinguishes her work at Endra. In this case, Bergman's cable-conduit algorithm ensures the optimizer evaluates routes against the actual building geometry, including the fittings that connect individual cable-tray segments into a usable routing network. When she later examined the downstream impact across real project models, she found that on one large U.S. hospital model, 32 of 33 fire-alarm loops relied on the connectivity provided by a single cable-tray fitting. Across the same model, accounting for the complete containment network changed the calculated fire-cable routing by nearly 10 kilometers, or 26.6%, illustrating how seemingly small geometric details can propagate into system-level engineering decisions.

The regulatory differences are only part of the challenge. Bergman says engineering firms often have their own conventions for how buildings are modeled, and many of those conventions were never designed to be interpreted by software. For decades, a building model only needed to make sense to engineers inside the same organization. A system expected to operate across firms and countries has to distinguish between requirements imposed by regulation and practices that developed because one firm has always worked a particular way.

"You can read the code," Bergman said. "What takes longer is understanding the decisions people make around the code and the conventions they do not think to explain because everyone around them already knows them." She describes that as the harder half of localization. Reading the formal requirements is only part of the work; learning how different organizations interpret and apply them introduces a layer of judgment that cannot be reduced to a settings menu.

That is why Bergman's role sits between customer insight, product shaping, and technical depth. She works directly with engineering firms, identifies where their workflows diverge from the assumptions already represented in the software, then turns those differences into product decisions and shipped behavior across the modeling stack and automation layer. The goal is not to collect an endless list of exceptions. It is to decide which differences reveal a broader pattern the system should be able to express.

"A special case is easy to add," Bergman said. "The harder question is whether the case exposes a more general rule. If it does, I want the system to handle that rule rather than solve only the example in front of us." She is more interested in finding the rule that explains the current case and potentially covers several more that have not appeared yet.

That approach becomes especially important in building design because the software is not producing abstract recommendations. Its output must fit into engineering work that eventually affects a physical building. Bergman's current responsibility includes encoding differences between Swedish, British, and American practice into a system that generates compliant electrical designs across whole projects. The same platform has to account for different approaches without losing the ability to produce something engineers can actually build from.

Independent coverage has identified Bergman's role in that international work. HackerNoon reported in July 2026 that she joined Endra to lead the work of making the platform native to American and British codes, standards, and design workflows, characterizing that localization effort as central to whether the product could travel beyond its home market. AEC Magazine separately reported in August 2026 that the adaptation of the platform to U.S. and U.K. codes and design workflows was being led by Bergman.

The coverage highlights a part of international software expansion that can remain largely invisible from the outside. An interface may look familiar while the system underneath it has to account for different assumptions about how professionals organize a design, interpret a requirement, or express the intent behind an engineering decision.

"Making software available in another country is easy compared with making it native to the way people there actually work," Bergman said. "If the engineer has to keep translating their process into the software's assumptions, you have not really localized the product."

Her longer-term interest in the U.S. market follows the same logic. Bergman says there is a limit to how well a team can understand American electrical practice from Stockholm. Codes can be read remotely, but conventions, tacit expectations, and the way projects actually unfold are learned by working alongside the people applying them. Endra is operating in New York, San Francisco, and London, and Bergman's goal is for the platform to work as a native tool in each environment rather than as a Swedish product with regional adjustments layered on top.

The high-rise lighting conflict is the kind of case that reveals whether that ambition is working. When Bergman encounters a new convention, she asks whether it should be handled narrowly or whether it reveals a more general rule the system needs to represent. Her job is to decide whether the software has encountered an exception, a new rule, or a signal that the underlying representation is too narrow.

That is what makes the work more than localization in the usual software sense. Bergman is trying to make one engineering system capable of understanding how different professionals arrive at valid designs under different rules and conventions. The platform may travel across borders, but for Bergman, the harder task is making sure the engineering logic travels with it.