Most of us don’t know what a geometric modeling kernel or solid modeling kernel is, but every CAD or design application uses one to do create shapes. It’s the underlying infrastructure that allows someone to do the math of CAD design. Just putting a hole in an object or blending lines or removing a corner requires calculating new edge curves that you wouldn’t want to do without the help of sophisticated algorithms. A kernel, which is a library of core mathematical functions that define and store 3D solid objects to support product modeling, makes it a relatively seamless process.
Interestingly, because building a kernel requires a lot of time — not to mention mathematicians and PhDs and computer scientists —the last geometry kernels to be built were rolled out in the late ’80s and early ’90s. Among the most popular is a the 3D ACIS modeler (ACIS), a geometric modeling kernel that was developed by Spatial Corporation and is now part of the French 3D design and engineering software giant Dassault Systèmes. The kernel is used in industries for everything from computer-aided design to 3D animation to and shipbuilding. Another is Parasolid, a geometric modeling kernel that was originally developed by a company called Shape Data Limited and is now owned by Siemens, which licenses it to other companies for use in their 3D computer graphics software products. Meanwhile, Autodesk began developing its own 3D modeling kernel called ShapeManager back in 2002. The companies’ virtual triopoly has enabled them to carve up what’s become a $50 billion between themselves, selling their licenses for steep prices and under terms that many find to be onerous. Indeed, despite occasional attempts by other companies to change the marketplace, none has yet managed to gain traction. But a three-year-old, Seattle-based company called Dyndrite that’s been quietly working away on its own kernel for four years, suggests things are about to change. “A big part of [what has foiled other upstarts] is they’ve gone head first after [these three giants], but trying to build something with all the features of a 30-year-old product is hard when you’re a startup,” says Dyndrite’s 26-year-old founder and CEO, Harshil Goel, who has one mathematics and three mechanical engineering degrees from UC Berkeley, including a PhD. Dyndrite has “taken a different tack,” he adds, “one that takes advantage of modern computer architectures, algorithms and languages and that’s completely GPU native.” Its kernel is build for the new world of 3D printing, says Goel. More specifically, he says, it’s capable of representing all current geometry types, including higher-order geometries, as well as can handle additive specific computations like lattice, support, and slice generation. In the process, says Voel, the kernel can reduce processing times from hours or days to minutes — even seconds. As Goel explains it, innovative manufacturing hardware has outpaced its software counterparts, and Dyndrite helps close the gap, Whether Dyndrite will break through is an open question, but early indicators are promising. The 15-person company just landed a little more than $10 million in Series A funding led by Gradient Ventures, Google’s AI-focused investment fund, which had also given the company earlier seed funding. Other investors in the round include Cota Capital and other earlier backers like Amplify Ventures, The House Fund, Flexport CEO Ryan Petersen, and former Autodesk CEO Carl Bass, whose relationship with Goel dates back nine years. They’d met when Goel was a freshman at Berkeley and Bass was on campus, talking with students and looking for new ideas to support. Says Goel, “I was a pure math major and spent all my time in machine shops. I met him one day and gave him a demonstration of what I was working on, and he was like, ‘Do I get to invest or what?'” He’s been Goel’s mentor since, says Goel, who says he aspires to have a similar career trajectory. “Carl just wants to be our worst end user by breaking everything,” he says, laughing.