Paul
Powers

Type a few words into Google and the whole internet answers. Hand a computer a coffee mug, a turbine blade, or a snapped bracket, and until recently it had nothing to say. Text was searchable. Shape was not. Paul Powers built the thing that closes that gap. Physna, the Columbus company he co-founded in 2015, reads the actual geometry of a 3D model and converts it into a unique code - a kind of fingerprint for objects. He calls it Physical DNA, and the company name is just that idea compressed: Physical DNA.

The pitch is deceptively simple. Engineers and designers can search a database not only with words, but with a sketch, a blueprint, a photograph, or another 3D model. Find the part that fits. Find the duplicate buried in a supplier catalog. Find the one component that broke and the seven others that share its weakness. Physna's own line for it sticks: it will show you where every tiny shard of glass belongs.

It is worth sitting with how counterintuitive that is. Two engineers can model the exact same bracket and produce files that, byte for byte, share almost nothing - different software, different ordering, different surfaces describing one identical object. Humans see the match instantly. Computers, until Physna, did not. The company's whole reason to exist is the gap between what a person recognizes at a glance and what software has historically been unable to reconcile. Powers built a translator for that gap, and he built it because as a lawyer chasing intellectual property he kept running headfirst into the wall it created.

What makes this more than a clever utility is the absence it filled. As Powers puts it, "There was no 'Google' for 3D, and no 'GitHub' for hardware engineers." Software ate the digital world by making it copyable, searchable, and version-controlled. The physical world - the part with weight and tolerances and broken edges - stayed illegible to machines. Physna's bet is that geometry deserves the same treatment text and code already got.

The detour that became the thesis

Powers did not arrive here by the engineer's usual road. He was homeschooled, which let an early obsession with science run unchecked, and entered Harvard at 16 to study astronomy and astrophysics. An exchange year in Switzerland handed him German. He took that language to the University of Heidelberg and did the least obvious thing imaginable with it: he earned a law degree and passed the German bar - reportedly with more than twice the required points - in a tongue he had picked up as a teenager.

He is cheerfully honest about how strange that looks. "If your goal is to get the most difficult useless degree you can think of, go to another country and take the bar exam in a foreign language, pass and then move back. That's essentially what I did," he has said. The point was never to practice law. It was leverage. He studied law to become a better entrepreneur, and he gravitated toward intellectual property.

That is where the company hides in plain sight. Lawyers already lean on algorithms to catch stolen logos and plagiarized text. Powers noticed those same algorithms went blind the moment the intellectual property was a physical product. You could not run a 3D part through software and ask whether someone had copied it, because nothing could match 3D data by its shape. He went looking for tools - "geometric search," "shape search" - and decided that instead of using them, he would build the company that made them work.

How Physical DNA actually works

Strip away the marketing and the mechanism is elegant. Most 3D files describe an object by listing surfaces and coordinates, which means two models of the same part can look completely different to a computer if they were drawn by different hands. Physna's approach reads the geometry itself - the relationships, curves, and proportions that make a bracket a bracket regardless of who modeled it - and encodes that into comparable data. Once geometry becomes a code, machines can do what they do best: match, rank, and learn.

The enterprise version of this lives inside manufacturers, where finding an existing part instead of designing a new one saves real money. The consumer-facing version is Thangs - a community platform that now holds tens of millions of 3D models, with search good enough that you can hunt for a design by shape and creators can monetize what they upload. One is the GitHub for hardware engineers Powers said the world was missing. The other is the Google for 3D. He is building both at once.

The two halves feed each other in a way that is easy to miss. Thangs is the public proving ground - a place where millions of designs flow in and the geometric-search engine gets tested against real, messy, contributor-made models every day. The enterprise product is where that same engine goes to work behind the firewall, scanning a company's own libraries for the duplicate part numbers, the near-identical components, and the quiet redundancies that bleed engineering hours. A consumer platform that doubles as a training ground for an enterprise tool is a deliberate piece of architecture, not an accident.

Powers has also leaned into letting creators get paid. Thangs introduced memberships so designers could monetize what they upload, a move that treats 3D models the way other platforms treat code, writing, or music - as work worth owning and selling. It is consistent with the intellectual-property instinct that started the whole company: if you can identify a design precisely enough to find it, you can also identify it precisely enough to protect and reward it.

Why Ohio, on purpose

Deep-tech orthodoxy says a company this technical belongs on a coast. Powers built it in Ohio - first Cincinnati, then Columbus - and treated the location as a feature rather than an apology. The funding answered back. In January 2021 Sequoia Capital led a $20M Series B, with Drive Capital alongside. Later that year a growth round led by Tiger Global added roughly $56M, pushing total funding past $86M. A geometric deep-learning startup in the Midwest pulled in some of the most pattern-matching investors in the business, which is its own kind of proof.

The operating philosophy

Powers talks about building like someone who has already absorbed a few hard quarters. He navigated the unexpected death of a co-founder in the company's early years, and the resilience shows up in how he frames the work. "Why do you care about yesterday? Last time I checked, today is a new day," he says. And, borrowing the old line and meaning it: "Whether you think you can, or you think you can't, you're right." His larger aspiration is cultural as much as commercial - he wants more people to treat failure as a catalyst rather than a verdict, so they take the kind of swings that built his own resume.

That posture also explains his comfort with the unconventional path. Since 16 he has run businesses of one kind or another - tutoring, translation, software - layering ventures on top of an education that zigzagged across three countries and three wildly different fields. He frames formal education less as a stack of facts and more as a long exercise in learning how to learn, and he is open about the fact that the hardest, strangest version of that exercise taught him the most. He did not collect a law degree in German because it was useful in the obvious sense. He collected it because the difficulty itself was the training.

There is a throughline in all of it. Astrophysics taught him to model systems too big to touch. Law taught him to find the hidden structure in messy disputes. Both, in the end, were about making something illegible legible. Physna does the same thing to the physical world - it gives shape a language, and then teaches machines to read it. The strange detour was the straight line all along.

The horizon he points at is bigger than a search box. He talks about digitizing the physical world - making real objects fully software-legible - which is the kind of ambition that sounds abstract until you remember what happened the last time something became searchable. Text got Google. Code got GitHub. If shape gets its own index, the downstream effects reach manufacturing, design, e-commerce, and every workflow that currently treats a physical part as a dead file instead of living, queryable data. Powers is wagering that geometry is the next thing to get a memory.

In his own words