Ep. 190 - Paul Powers, CEO of Physna on Machine Learning, 3-D Data, and Building Startups in the Midwest

On this week's episode of Inside Outside Innovation, Brian Ardinger, Inside Outside Innovation Founder, sits down with Paul Powers. Paul is the CEO and co-founder of Physna. They talk about innovation in the manufacturing space, 3-D data, trends Paul is seeing from the CES conference, and building a startup outside the Valley in Cincinnati, Ohio.

Inside Outside Innovation is the podcast that brings you the best and the brightest in the world of startups and innovation. I'm your host Brian Ardinger, founder of insideoutside.io, a provider of research, events, and consulting services that help innovators and entrepreneurs build better products, launch new ideas, and compete in a world of change and disruption. Each week we'll give you a front row seat to the latest thinking tools, tactics, and trends, and collaborative innovation. Let's get started.

To read the interview transcript, go to insideoutside.io

Interview Transcript

Brian Ardinger:  Welcome to another episode of Inside Outside Innovation. I'm your host Brian Ardinger, and as always, we have another amazing guest. Today we have Paul Powers. Paul is a Forbes 30 under 30, a graduate of Heidelberg University with a law degree.  He is an astronomy and astrophysics alumni at Harvard. He's a serial entrepreneur, and his most recent startup company is Physna, which he started in 2015. Welcome to the show, Paul. 

Paul Powers: Thank you. 

Brian Ardinger: You've got a pretty extraordinary background.  I wanted to have you on the show for a couple different reasons. One, because you're a young founder out there in the world building some interesting things.  Your company Physna is in the manufacturing space, and we haven't had a lot of folks on the show to talk about manufacturing innovation.  I thought it'd be a really good opportunity to start the conversation with, tell us a little bit about Physna and what does it do. 

Physna and 3-D Data

Paul Powers: So Physna is short for physical DNA. And what we do is we take three-dimensional data and we normalize that down into something that software can actually read. And we help to bridge the gap between software applications that are tech space of two dimensional, and the real world essentially, which is obviously physical and three-dimensional. We do that through a series of proprietary algorithms and we applied machine learning to our technology so that we can actually not only break down and comprehend what we're looking at, but also make predictions about how humans might classify that, that might be used for, how you might make it, what are my costs, how's my performance, certain situations, et cetera. The most common use cases for the technology are to use it to help with engineering, to speed up the process so that you're not redesigning things from scratch and helps you make predictions about what you're trying to design and speed it up.

It helps in procurement by understanding what options you have. What suppliers might be able to provide the components that this thing has inside of it and who might be able to manufacture it, at what costs, et cetera. And then under the manufacturing side, understanding how to manufacture those, how it might turn out qualitatively, predicting quality, and a number of users out there who use it for a couple of other things marked miscellaneous use cases. We do have some work that we do together with the military, for example, to identify parts in the fields that aren't necessarily even a CAD model at that point. They can use AI or an image or even a 3-D scan to figure out what something is and more information about it. 

Journey to Finding Patent Problems

Brian Ardinger:  Tell us a little bit about how you got started in this space. My understanding is you started with a law degree and a law background. How did you get to designing software to attack the patent problems and everything else in the physical world? 

Paul Powers: It's not obviously a very direct line between those two things. What happened was I studied law because I wanted to be an entrepreneur and I thought that might be given me an edge or it might just be a different way of looking at starting a business. I focused on intellectual property. That was the closest thing to technology it felt like. It was cool. You got to see a lot of neat things, but I knew that being in that field bet, it was really easy to find a patent, like, you know violations of people's logos or music, texts written as a book or whatever.  It's easy to find digital copies of that, that are not legally obtained. But as soon as it comes in those 3-D models, it was very difficult. We can never really predict violations that might be about to occur, and that really is because it was hard to even search for a 3-D model with a 3-D model or with other input.

They're hard to identify. They're hard to understand. There's so many file formats out there, and you certainly can't really find it without like a perfect match of a 3 D model, it seemed. We started the company for that because there's so much cost to that problem. It's trillions of dollars annually and global loss for patent violations. And we thought if we can tackle that problem, that a benefit to society, not just because you can make more money off of your ideas, but also because it helps promote research and development by lowering the likelihood of the theft of your IP. We launched the company, we tried out everything in the world we could find that had geometric search or shape search or anything like that that we thought would be relevant.  And we tried out a lot of stuff and everything was extremely disappointing to us.

Everything in the market, we always keep an eye to trying out other tools, but they were very disappointing because they didn't really do what we thought. They weren't actually breaking the stuff down into 3-D. We found a way after a lot of time, but eventually we figured out our own way to break down through the models and to truly break them down so that you could find parts with a subsection. For example, let's say you have a screw and you want to use that to find a machine that it goes into. You can do that, or if you have half of a part, you can use that to find the rest of it. You can identify what's inside of the part even if you don't have data like this is the parent file these with the children's files, but you don't have to have that.  We can figure that out. 

Once we had that, we went out to conventions and started telling people about this technology, and very quickly we started hearing about all these other issues that existed that, frankly, I had no idea existed, right. That engineers, manufacturing, mechanical engineers, electrical engineers, anyone who engineers something physical, you know, not a software engineer, their productivity is only 20% of what it should be. If you compare how effective software engineers are compared to engineers of physical goods, that's over a five to one ratio actually. That's because these tools are missing. They redesign things from scratch, etc. We also found issues in quality control and inspection automation, and even in healthcare, and all these other areas. And we got overwhelmed and realized that, wow, the reason that we're finding so many issues and people are coming to u...