Chasing Butterflies

This podcast episode of AM Insider features a discussion with Brent Ewald, a software development manager at Autodesk, about his diverse career journey within the additive manufacturing space. The hosts, Justin Hopkins and Dustin Kloempken, explore Ewald's transition from mechanical engineering and R&D in 3D printing at HP to software development. Their conversation examines the evolution of software in additive manufacturing, the importance of data analysis, and the future potential of technologies like implicit modeling and standardized file formats. They also touch upon the skills needed for those entering the software side of the additive industry and briefly discuss the emergence of AI in this field.

• Career Navigation in Additive Manufacturing
• The Crucial Role of Software in the Future of Additive Manufacturing
• Advice for Individuals Interested in Software in Additive Manufacturing
• Geometry Representation in CAD and Additive Manufacturing
• The Impact of Artificial Intelligence (AI) on Additive Manufacturing

Transcript

S2 EP3 Brent Butterflies

Welcome to AM Insider, a production of our own. In this podcast, we will have a series of
informative discussions tackling the adoption of additive manufacturing, answering those
burning questions and swapping experiences. Learn from experienced individuals on how
innovation can push the boundaries of what is possible.
Hey, and welcome to the podcast. My name is Justin Hopkins, and with me here as always
is Dustin Kloempken. Let's get inside AM.
All right. Well, welcome to this next edition of AM Insider. And today we have a very special
guest that we've worked with in the past, and it's going to be very interesting to hear what he's
doing now and get his insights into the realm in which he works in.
And, you know, Dustin, from your past experiences working with this individual, what can you
kind of give the audience a little background on? Yeah, so I'm excited to talk about this with this
person as well today, because to me, this person in particular, he comes across as very easy to
talk to and to relate to. But if you pull back the onion, a couple of layers, you realize that he has
just a huge depth of knowledge and experience and understanding and quite a few different
topics and sectors. And so what I've really appreciated with working with him over the years is
he's able to kind of meet you where you are today with knowledge or experience, and then take
you to what he knows in a way that makes sense and is intuitive.
So this person has a very deep experience, I would say, in the software space, and I'm kind of
excited to see where the conversation will go today. Yeah, absolutely. And I know he has a lot of
experience outside of software too that we used to talk about, and it was rather interesting.
So without further ado, let's bring in Brent Ewald. Welcome to the show, Brent. Thank you,
Dustin.
Thank you, Dustin. I'm happy to be here. It's been a long time.
I haven't talked to you in quite some time, and I'm looking forward to doing this conversation
today. And how have you been? And can you give the audience a little bit of background about
where you are now and what you've done throughout the years? Yeah, absolutely. So I'll start
with right now.
I'm a software development manager at Autodesk, and everything I say is my opinions, not that
of Autodesk. I don't know. I feel like I should say that.
My background, let's see. I've done startups in the software space, and then I also was at HP
for, I think, what turned out to be seven years non-contiguous, mostly in the 3D printing space.
So I was kind of really early on in 2014 and 2015 when they were developing the tech.
I was working on R&D, the printer design there. Actually, a lot of people might not know this.
Well, the people who I work with, obviously some people who don't know me, never know this
stuff, that my background, I have two mechanical engineering degrees.
I do not have a software degree. That's just kind of learned by necessity, and then now it's what
I do. I mostly have worked in the additive space, though, if you go back far enough, I've worked
in automotive, power engineering, all sorts of interesting fields.
It gives me a lot of insight into how things are actually done. But yeah, that's me, and I'm just
always eager to learn, and eager to try new things. Okay.
And so from your journey, and like you said, you worked in the beginning stages at HP on
printer R&D to software, and it's like like you mentioned, you didn't have the background in
software, but tell us a little bit about that journey, like how you got interested into that, what
got you into that space, and how did you get there? Right. So at HP, when we were developing,
I worked on the color printer, may it rest in peace, and it was such an interesting journey
because this was the first time we had to think about bd color propagating off of surfaces,
because really all the file formats defined color on surfaces, there's all these crazy
considerations that like were new to HP, but also fairly new to the industry. And so that was, it
was one of the best experiences I had.
But the reason I got into software from there was because these machines were generating,
like, they could generate 10 gigabytes of data in one build, you know, some crazy number. And
we had to, like, we were developing complex systems, like developing a furnace in a box, plus,
you know, a microwave, I don't know. But you put all that together and these complex systems
printing out tons of data, you have to be able to organize it, wrangle it to understand it and
program.
And really, I got started in programming from the data side. What kind of data were you
capturing? Was it just feedback from the system or what was the data I'm telling? Yeah, so this
was actually my favorite part about it was that the data came from every servo, like every motor
control, the lamps, the heating elements and all this stuff, even a thermal camera. And so we
were, you know, the end thing that was really weird was, let's say you're developing this printer
and it's not done yet.
So you have defects and you look at a part that was printed over the course of, I don't know
how many hours. And then you look at data and you try to figure out what happened here that
made it do this. And that was just so much fun.
So yeah, that was ultimately like it was everything. Everything was in there. So something that
we've talked about with some other speakers in the past a little bit is, you know, navigating
your career within the additive space.
And so I'm kind of curious, you know, given that you started in R&D and then you came onto
the AE team and then since switched into software, I mean, those are pretty big switches. I'm
kind of curious, you know, what did you realize were your motivations to kind of go in that
direction? And was there anything in particular that would have been useful to know about
before you made those changes? Because of course, there isn't exactly a perfect career path in
the space, you know? Yeah, it's also because the industry is so new. It's unbelievable how new
this industry is when you compare it to almost anything else.
But so the big thing I think I would, you didn't ask it this way, but I'll state it this way, is I would
have been an AE first. I think that would have been so much more helpful if everyone who was
designing that 3D printer had been an AE in the 3D printing field ahead of time. You just don't
understand the reality of things if you're in this lab condition versus talking to customers on
their sites and understanding their problems.
Yeah, that kind of brings up, it's always a big hole. I think at every OEM is R&D isn't necessarily
in direct access to the end customer. So there's this missing link as to like, kind of like what
you're saying, what you're doing in a lab as opposed to what's actually going to happen once it
gets out in the field.
Yeah, there are so many assumptions that we're like, oh, in hindsight, we could have done that
differently. But that's with everything. Yeah, so if I were to change anything about the career
trajectory, I would have done AE first.
Now, the other thing you talked about like R&D, AE, and software now. Yeah, I wouldn't advise
that. Like, I wouldn't tell anyone to do all those things.
But that's hindsight, right? It took to learn software, I not only like worked a job, and you guys
know, I'm a family man, I have two kids and a wife, and I'm just always busy. But then I took,
what is it? I got C++ for dummies, the complete volume, and it is maybe three inches thick. And
I just read it cover to cover while doing practice examples at night.
And in the end, I can now do C++. But you know, that's I wouldn't advise that for anyone. No, if
you know what you want to do, just do it, but maybe be an AE first.
And what possessed you to do that? Like, why did you want C++? Well, so at first, I was learning
Python out of necessity. This was when I was at HP, wrangling all this data, and I got pretty
good at that. And then I needed to do C++, because I needed my code to be more performant.
That was basically it. I started dealing with projects like giant meshing projects and Booleans
and all these crazy things. And I was like, if I did this in Python, it would not, it would take
forever.
Just a quick background here on the difference between the two. There's many, many
differences. But this a significant one that I'm talking about right now is that Python is an
interpreted language versus a compiled language.
And I know that means almost nothing to most of your audience, but interpreted languages,
there's scripts, you can just type some and run it. Compiled, you have to compile it, and there's
optimizations that take place. So it's a matter of what are you trying to achieve? And say, I'll just
comment here that for people who are thinking about their career path and journey within
additive, the big thing that I've really seen, like with Brent, for example, is the people who do
move around and make those jumps, they just find a way to make it happen.
If it's reading a big book or taking classes or something like that, because right now there isn't
really a lot of structure when it comes to that kind of stuff. And so if you're kind of new here,
listening in from outside of the industry and thinking of jumping into additive, just know that it
is kind of the wild, wild west, kind of like you're, you're highlighting here, Brent. Absolutely.
The other thing I did, I did though, is I had to find someone who would take a chance on me. So
that's the other thing, because, you know, hiring a software engineer with no software
background is kind of a tall order. Hey, it worked out.
Look at this. Yeah, exactly. No, I really enjoy this stuff.
And so tell us a little bit about some of the projects you worked on on the software side. Cause
you mentioned using Python, like what were you doing with Python and then how did that
involve or evolve into what you're doing now? Right. So when I was doing data wrangling and
stuff is what I'm going to call it.
I was, the real purpose was like, we were trying to figure out early indicators, what system is
failing? Cause when you get a failed print, it's, you don't always know why it fails, right? You,
you see something got strewn about the build. So you say, why did that happen? And what
controls this? And so I was just parsing vast amounts of data and then putting them into a
database and then using analysis tools to like BI tools, like, you know, all those business
intelligence stuff people use now for dashboards. And I feel like it's been overused, but how did
it evolve to this was more or less, I started getting into the computational geometry aspect of it.
And actually that's something that has a ton of value in additive. It's one of those things where
like the hardware introduced a new capability and then the market doesn't have a way to use it.
And so that's actually the interesting part about software and additive is that you're filling this
gap that didn't exist.
Yeah. So kind of expanding on that, you know, one of the reasons why I'm really curious with
talking with you today is I get asked a lot, Hey, what is the future of additive? What's going to be
the next big breakthrough? You know, what do you see is up and coming? And my own
personal opinion is having been in this industry for over 10 years now, a lot of the printers have
gotten pretty solid. The materials have gotten pretty solid.
The printing process is getting more intuitive and easier to do, but where some of the gaps still
are, in my opinion, is in software. And so I really believe that some of the next big
breakthroughs will be coming through in that space. And so one of the questions that I'm
curious about now that you've been in the software space with additive for a little while now,
I'm curious, what you were seeing is coming and what's what gets you excited and wants to
make you get up and get going in the morning.
Well, I think the thing that it's integrations is really the big thing. And I know this is like
cooperation and the ecosystem and stuff like that, but that's an interoperability. So this is the
stuff that like the additive industry didn't do for the longest time.
And it maybe it's because there is so much venture capital dumped into it or all these other
things where people are all, there's all these startups and there's no cohesion in the ecosystem.
But I'm on the 3MF, I'm a steering member of the 3MF consortium through Autodesk. And like
one big push we have is interoperability.
And so this is stuff like we actually released a volumetric or an implicit file format. So now we
can communicate implicit geometry without having to generate, you know, a hundred gigabyte
meshes or something crazy. So it's that kind of stuff that just changes the game.
So implicit modeling is always a big thing, but then there's other ecosystem type things that I'm
working on that are very, very important. So the ntop interop, you can now import ntop files
dot implicit files into Fusion. And that's just really cool.
That is really cool. And to go back to what you mentioned about the voxels and the 3MF file for
the implicits, that seems like something that's new to me, because 3MF from my perspective for
a long time is a better file format than STL, but it hasn't evolved in the way that I think people
expected it to. But you kind of mentioned something to me that is different than what I've seen
recently.
Right. So I think the way 3MF is conceived is that it's this format. It's not meant to be just a
mesh format.
It's a, you know, 3D manufacturing format, 3MF. And there's the core, the core thing of 3MF,
but then there are extensions. And these are like things like, have you heard of the beam lattice
extension? Yes.
All right. So that provides a language to describe a lattice just by points and connections and
diameters, right? Or radio. I don't recall which, but that's not the point.
It provides another language to communicate geometry that would otherwise be, you know,
really unwieldy. So, but that's an extension. So I think the real challenge we've had in the, for
3MF is getting people to uniformly adopt the extension, because who, do you know anyone
who uses the beam format? I know a couple people, but I don't know if it's probably.
Yeah. I haven't come across it recently. Sometimes you might with people doing line structures
and stuff, but not, it would be very rare that I would see a file come to me that way.
Exactly. And that's because, you know, we as the software industry need to do a better job of
standardizing and all adopting those extensions to make them useful. If I can generate a beam
lattice, that's great.
It's a compressed format, you know, it's essentially a massive compression while preserving
perfect cylinders or what it cones. But what it doesn't do, if I can't share it with anyone, that's
pretty useless, you know, how can I print it? So it's that kind of stuff that's really essential. And
so we have that beam lattice.
There's an implicit version of it. There's a Boolean version that's either coming or out. I'm not
certain, but yeah, there's tons of extensions, but until we get broader commercial adoption, it's
hard.
It's a great point. And just kind of to pull a little bit of a thread there. So if there are people
listening in that are looking to get into the software space within additive, is that something
specifically that you'd suggest that they look into and learn more about? Or are there other
things that you think they should study and focus on? I know it's a fast moving ship, but you
know, I mean, people I'm sure people ask you on occasion, hey, if I get into additive, what
should I learn? What should I study? What should I look into? And I'm just thinking for people
who are interested in software in particular.
You know, what's funny is there's enough problems to be solved that I would say, what are you
interested in? I guarantee you could find a problem to be solved for any skill set or any interest.
I came across this really cool company the other day. I think it's called Euler or E-U-L-E-R.
And they do, they do it's some sort of magic, maybe AI, I don't know, but it's like for quality
control of prints. And you know, video of the, and it's for metal printing, I think, for SLS. But you
just, it analyzes the builds visually, and then points out where there might be print defects.
So if you like vision code, that's great. If you like quality metrics, if you like material testing, like
you can come up with any sort of, I don't know, if you like playing Tetris, I'm sure there's a way
to put that into a software thing that you can solve. So the added, it's so rich.
You take any problem and there's probably a software solution to it. So let me rephrase it a
little bit. If there were some core skills or knowledge sets that they should learn, is there
anything that kind of comes to mind? No, it's just moving so fast that I don't know if I like, be
comfortable in ambiguity, really hone your skill of understanding the problem.
And yeah, I don't, there's no like, you need to know how to code. That's like the base, the bare,
the base entry point. But other than that, be curious.
And you'll get very far. Is there any coding language in particular that's the most used or
common that you've seen? I see people more and more using Python, and then forms of
JavaScript. And then the thing is, if you talk to like hardcore CAD people, they all say C++ or, you
know, some of these old things I've seen are written in Fortran.
Like, sorry, I'm just throwing out like languages here. If you don't know what these are, like
Python is probably the easiest. It's like a scripting language.
JavaScript is probably even easier, but it's really limited in terms of libraries, but, not libraries,
but CAD geometry stuff. And then Fortran is just incredible. Like that's the stuff.
I don't know, I think it's like C, it's as old as C. I'm making some of the software. It's very old
language and nothing's really written in it. Well, that's still helpful.
It at least gives people a direction if they're looking to, you know, either pivot into software or
just start altogether. Yeah. Well, the other, okay, I can give a piece of advice is make it useful to
your everyday things.
So if that means Python, that's great. Like I don't look down or up on any link, you know, like
languages are languages. They're just tools.
Basically, if you can apply it to your day-to-day work, that's a great thing because you'll develop
it. You'll keep moving the needle forward. So it's just consistency.
And on that same kind of token, you mentioned these different languages and being at
Autodesk and working within Fusion. Fusion's kind of bringing together a lot of these different
modeling types. And there's just as many modeling types almost as there is coding languages,
right? So could you give everybody kind of a little bit of background on those different types of
modeling? Yeah, I will say I'm not like a super user of Fusion.
You know, I work on the back, but high level there, if you look at it, there's solid mesh. I know
we have a sheet metal design type, but, and then there's a form where you can do like sculpting
and pretty interactive stuff, kind of like sub D. We also have a volumetric inside the product
design extension. And the other thing that I think is really interesting with Fusion and how we
do these integrations is our API.
So if you have something you want to do in Fusion, quite often, if it's not in base Fusion, but you
like everything else, you can just add it on and write your own little script that does it, add
buttons to the command and everything like that. But if I were to back up a bit, let's talk about
geometry. Generally, geometry comes in many different forms.
Justin, maybe you can name a few of these or Dustin. I could throw a few out. So you have B
reps, and then you have voxels, and then you have point cloud.
I mean, the list can kind of go on and on. Dustin, do you want to add to that? Well, throw mesh
out there because that's the most common, but go ahead. Yeah, you know, honestly, this may
sound weird to the audience.
I've been in the industry a long time, but this is a sector that I haven't really put my thumb on
the scale much because it's never really become that much of an issue for me to really learn
more about. Now, I know people like Brent has really dug into it. And I mean, there is so much
to know with this that I'm just not as experienced when it comes to that kind of thing.
And so to me, I think it's important for the audience to see like Justin and I have both been in
the industry a long time. He happens to know a bit about it. Brent knows a lot about it.
And I know very little about it. It just kind of shows that, you know, if you're in an application
engineering role, which we all were at one time together, we can have all very different types of
knowledge and experience. And that's okay, because there is a need for it, kind of like what
Brent mentioned.
But I digress a little bit Brent, I'll let you jump back to it. Yeah, so I have a high level list like ever
sketched out in front of me here, you got most of them. So point cloud, and this is like, well, I'll
go into that in a second, but point cloud, B rep, mesh, voxel, implicit, there's many more, but I
think at a high level, that's a good, a good scope.
You can also think like tool paths are really like lines that fill a volume or something. You can
think of it in many different ways. But the reason why I asked that is because, you know, I like
to communicate this stuff through analogies.
So B rep is called, it stands for boundary representation. So it's like defining a shape by its
outside. And that's an easy way to think of it, right? And there's a few different surfaces and
stuff, but nerves is like a common one.
There's a joke, what does nerve stand for? Nobody understands rational B splines. I have not
heard that one before. Well, only a very small portion of population would get that joke.
But yeah, that's what it stands for. And so that's like your old school CAD. Like when people
were like writing CAD kernels, they, they always dealt with B reps.
And then what happened was 3d printing was like, well, I need to get explicit, right? I need to, I
need to have, well, actually graphics was a big driver of mesh too, for a while in video games,
animation and stuff, but VR and all. Ah, there it is. Yeah, because VMR, from my understanding,
I was told, because I worked with the RML a lot, it was used in early game design, and the 3d
industry repurposed it and was able to use it for color printing.
Yeah. I mean, it's amazing how this stuff just builds off of whatever was there before. Very
organic that way.
But moving forward from that. So there's mesh and B rep. So they're different meshes, like
actual flat faces, you know, you can't have a curved mesh, right? But if I were to make an
analogy, voxels are like the mesh of implicit, I'm gonna let that sit for a second.
Well said. Okay, good. Okay, good.
Yeah. So like, you can make analogies between all these geometry types. But yeah, you have to
master them to understand what it is you're doing, especially when you get to like, workflows
and automations and complicated stuff.
And also, one thing I like to think about when I talk about this stuff is what, what type does the
printer in the end care about? And for those, those people who are listening in completely lost,
which I'm sure some of you might be, I think the important thing is, at least from what I've seen
working with Brent over the years is, you know, all this stuff generally probably doesn't matter
to the average user of the additive manufacturing space, unless they run into a situation or
some kind of problem. Because what I saw you do a lot of times Brent is there'd be some really
challenging problem. And you'd say, I don't really know how to solve it.
But let me dig into it. And you would dig into it and literally teach yourself how to code like you
mentioned, in order to solve the problem. And I think that's kind of the takeaway that I keep
hearing from you is, you know, where you study and focus your time, if you want to get into the
space, only matters if you can use it in your day-to-day world.
That's a great way to put it. Yeah. And again, I'm not advising people make that many jumps
and do that much breadth.
It's exhausting a bit. But it is, I mean, it is really interesting to see it at all those different levels.
Is there anything you guys are wondering about like big things in the industry trends?
Definitely AI.
Dustin always loves to cover the topic of AI, but like how that's changing our space in additive
manufacturing, that we see text to geometry, but what other things might happen? I mean,
you'll see more and more as it will probably get integrated into hardware or data sets, like you
said, of what's happening afterwards and how to interpret certain things. For the most part, I
think most people look at like, what can I text to geometry? Absolutely. And in fact, I think HP is
breaking, you know, the latest news on that stuff.
I saw your, I was at, I wasn't at form next. I got sick right before, but I saw your, you know, HP's
announcement and all those texts too. Like it was making the rounds on LinkedIn.
It was like a donut. Did you with the sprinkles? I really liked that. But if we peel back to the
onion a little bit, yeah, like generative AI is, AI in general is like scaring people right now.
Cause who knows what is going to fall next or, you know, what's going to come out of it. But I
have a little, a thought experiment. Why, why do we think LLMs and like image gen like Dolly,
why were they the first things to crack? And by crack, you know what I mean? Like, like they, oh,
we figured it out.
Boom. Generative AI. Why do you think I have my opinion? I'm curious.
Are you asking why they came first, quote unquote versus like there's mesh generation coming
out now, or there's like AI agents or all these things coming out now. But why do you think
LLMs and, and Dolly were like the first things? Cause if you think back to like 2000, 2022, 2021
ish, when all this stuff was just like completely new, it was LLMs that really, and Dolly that blew
the world blew their mind. So I can actually speak to that a little bit.
So around the time chat GPT came out, I was taking a course and someone in that course was
from one of the big tech companies, just to keep them anonymous. I'm not going to say
anything. What he said, what he told me, which I found very interesting, and this is just one
source.
So take it as it is. He said, look, we were working on LLMs for a long time internally at our
company and we could have launched it to the world, but we found that it made too many
hallucinations. And we felt that ethically, it is not wise to release something like that into the
wild, into the world when people could rely on bad data.
Well, then, then open AI came out with their big flagship product. And as we all know, there
were lots of hallucinations. I would say it's much better today than it was then.
But, um, the person at this tech company, they said that generally they, they thought that was
rather reckless because they, they thought, okay, you're going in to try to get fast market share
and you did, but that's not good for the public good. So I thought that was kind of an
interesting story that I hadn't really heard or considered before then. Oh, not the direction I
was thinking you'd go, but a very good point.
And yeah, that's interesting. Justin, is there anything in particular? I mean, from my side, I
mean, if you're talking about language and, and photographs or pictures, they're just more
common to the masses, right? So it's going to be adopted more easily just due to the amount of
people that will consume it over something that might be in a smaller field, such as ours. Yeah.
And so that's the direction I was thinking. In fact, if I were to flip the statement and say, there is
significantly more data available for text and image to train such a massive model. And so it's
exactly what you said, but I just would frame it a little differently because how many CAD
designs of gears are there that are meaningfully different, that are available that you can just
get, I don't know, maybe in a thousand versus text, you can just find anywhere.
You go on the internet, all you get is text or images. And so it's like the digitization and publicly
available amount of information is what makes it interesting or at least capable. Maybe with
some dubious ethics, but it's the information that really enabled those two.
So if we're talking about like 3D AI,
This file is longer than 30 minutes.
Go Unlimited at TurboScribe.ai to transcribe files up to 10 hours long.