If you’ve ever had the pleasure of putting together furniture, you know how grueling it can be to squint at diagrams and guess which screw goes where.
Building submarines and jet engines is certainly more complex than assembling a bed frame, but when it comes to assembly instructions, even the most advanced manufacturers sometimes face the same headache. One reindustrialization-pilled startup thinks it has the fix.
“You know that paper instruction book that tells you how to build something?” Fil Aronshtein, Dirac’s cofounder and CEO, asked, referencing a harrowing experience with an Ikea bookshelf. “It turns out everything around us also needs assembly instructions.”
Aronshtein says BuildOS, Dirac’s main product, is a software tool that uses AI to generate assembly instructions for manufacturers. The startup raised nearly $11 million from Founders Fund and Coatue Management — and struck a new partnership with industrial manufacturing giant Siemens.
Typically, the workflow for building a piece of hard tech goes something like this: “Some dude gets a CAD file over email, takes hundreds of screenshots manually, figures out what order to do the assembly, then takes all those screenshots and throws them together into a 100-page PowerPoint,” Aronshtein said. “It’s super manual, and super tedious.”
Instead of relying on humans to stitch together those convoluted instructions, the 18-person startup says it uses artificial intelligence to determine the order of assembly. (Aronshtein declined to disclose whether the company uses a foundation model or trains its own.)
In 2023, Aronshtein cofounded Dirac with Peter Weiss, a college friend who briefly worked as a software engineer at Amazon. Aronshtein had a short stint at Northrop Grumman. The two met as electrical engineering undergrads at Johns Hopkins, where they both earned master’s degrees in robotics.
With the Siemens partnership under its belt, Dirac is integrating BuildOS with Teamcenter, Siemens’ engineering management software. That integration will allow Dirac to work more closely with Siemens’ customer base, said Tomas Klausing, a director of technology partnerships at Siemens. (Aronshtein wouldn’t name customers but said Dirac is working with “some of the coolest companies on the planet in aerospace, defense, automotive, agriculture, construction, machinery, and maritime.”)
Dirac isn’t the only company trying to modernize the shop floor. Tulip, a manufacturing software startup that last raised $100 million in 2021, is also making software for machinery with a platform that aims to track factory production and equipment assembly.
‘Capacity to Build’
Much of Dirac’s origin story is tied to Silicon Valley’s recent push to “reindustrialize” the US, a movement aimed at reshoring production and keeping pace with China’s blistering industrial growth.
Related stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
“The foundation of the West was built upon our capacity to build,” Aronshtein said. But during his time at Northrop Grumman, he was struck by how “ridiculously archaic all of the infrastructure was on the manufacturing side” — something he wanted to change.
That insight came as funding for defense tech startups was surging. Trae Stephens, a partner at Peter Thiel’s Founders Fund and cofounder of Anduril, bet on Dirac after seeing firsthand how much the defense industry hinged on US-based production.
Getting America’s factories up to speed has become a recent obsession among venture capitalists: “The investments that we make become consensus,” Stephens told Business Insider about Founders Fund. “The most pro-democracy, pro-West decision we can make is to bring automation and robotics into manufacturing.”
For Thomas Laffont, cofounder of Coatue, the Dirac investment was also about leveraging software for hardware.
“There’s a broad movement toward rebuilding infrastructure. The obvious issue with that is, we need to do it in a way that is cost consistent,” Laffont told Business Insider. “Then the question will be, ‘How do we make reindustrialization work within an economic constraint?’ We think it’s AI and software.”
