Being an Engineer
Being an Engineer
S6E40 Matt Puchalski | Vision Inspection, Autonomous Vehicles, & Graduating Y Combinator
Matt Puchalski is a roboticist and entrepreneur shaping the future of automation in manufacturing. As the founder and CEO of Bucket Robotics—part of Y Combinator’s Summer 2024 batch—Matt is building next-generation 3D vision systems designed to supercharge flexible manufacturing environments. His company combines high-quality monocular cameras with edge computing to enable real-time 3D perception, simplify integration, and generate meaningful metrics across production workflows.
Before founding Bucket Robotics, Matt spent over half a decade developing and deploying Level 4 autonomous vehicles at Argo AI, which was acquired by Ford to form Latitude AI. His journey in autonomy continued at Latitude and then Stack AV, where he served as a foundational engineer during the company’s stealth phase. From leading test track operations to engineering vehicle reliability processes, Matt played key roles in launching multiple AV platforms on public roads.
He’s also an inventor, credited with patents ranging from indoor localization to dynamic data mining for autonomous systems. Alongside his startup leadership, Matt is a venture partner at Pioneer Fund, helping other early-stage founders navigate the startup ecosystem.
Matt holds a B.S. in Electrical and Electronics Engineering from Georgia Tech and brings a unique blend of hardware expertise, startup grit, and large-scale system thinking. Whether deploying AVs in Munich or building user-friendly tools for manufacturers, Matt is passionate about making robotics practical, accessible, and powerful.
LINKS:
Guest LinkedIn: https://www.linkedin.com/in/matt-puchalski/
Guest website: https://www.bucket.bot/
Aaron Moncur, host
Click here to learn more about simulation solutions from Simutech Group.
🚀 Join Us at PDX 2025! 🚀
PDX 2025 is the Product Development Expo designed for engineers who want hands-on training from industry experts. PDX focuses on practical skill-building, cutting-edge tools, and real-world solutions.
📅 October 21-22, 2025
📍 Mesa Convention Center, AZ
🔗 https://reg.eventmobi.com/product-development-expo-2025
About Being An Engineer
The Being An Engineer podcast is a repository for industry knowledge and a tool through which engineers learn about and connect with relevant companies, technologies, people resources, and opportunities. We feature successful mechanical engineers and interview engineers who are passionate about their work and who made a great impact on the engineering community.
The Being An Engineer podcast is brought to you by Pipeline Design & Engineering. Pipeline partners with medical & other device engineering teams who need turnkey equipment such as cycle test machines, custom test fixtures, automation equipment, assembly jigs, inspection stations and more. You can find us on the web at www.teampipeline.us
Matt, hello and welcome to the being an engineer podcast. Today's guest is Matt Puchalski, founder and CEO of bucket robotics, a Y Combinator backed startup developing advanced vision systems for flexible manufacturing with deep experience in autonomous vehicle integration from companies like Argo AI latitude, AI and stack. AV, Matt brings a unique perspective on deploying intelligent robotic systems at scale. Matt, thank you so much for being with us today.
Matt Puchalski:It's great to be on I'm really excited to talk to you, Aaron, this is really exciting.
Aaron Moncur:Awesome, awesome. Yeah, so Matt and I connected a few weeks ago. A mutual friend Ian mccatherne introduced us and shout out to Ian out there. He's been on the show a couple of times now. Rad guy, just so smart, knows so many things about engineering, all
Matt Puchalski:descriptors that I would use for Ian. Ian is the best. He's part of our new hire onboarding plan, which is, go spend half a day with Ian and then buy him coffee. The guy's great.
Aaron Moncur:He is, yes, he is a gem, a diamond in the rough, for sure. All right, so yeah, we connected a few weeks ago, and Matt was telling me about his company and robotics they're working on. They just landed a big contract, which is super exciting. And I thought, man, we got to have this guy on the show and share about all the cool things that he's doing. So super grateful to you, Matt, for taking some time out of your your busy schedule, and let's, let's dig into it. So first question, what made you decide to become an engineer? Oh my
Matt Puchalski:gosh, this is such a great question. I my parents actually sort of begged me not to be an engineer. My dad is an environmental engineer, and when I was in high school, maybe a freshman or sophomore, as I was going further and further into math classes and really enjoying that, I had some incredible mentors. I told my parents, hey guys, I think I want to be like dad. And they were like, please. No, be a lawyer. Do anything else. No, seriously. Why?
Aaron Moncur:Really? Why were they against it? Tongue in cheek, or
Matt Puchalski:do this? No, it was definitely tongue in cheek. I'm the oldest so, you know, they were learning like what to do, kind of thing, you know, I in in high school, I had a very diverse set of interests. It was between electrical engineering and Latin. Were my two different true passions. I I'm still consider myself a Latin scholar. That was fun. You know, got a perfect score all four years in the National Latin Exam. But tell
Aaron Moncur:me a little bit about that. I mean, that's unique and interesting. Why Latin? Where did that come from? Yeah.
Matt Puchalski:So I grew up in New Jersey, in the smack middle freehold New Jersey, the part that Bruce Springsteen sings about leaving, but it's a great town to grow up in, and I was just a giant history dork. I really gravitated towards learning about, you know, European history. We were talking about Istanbul, just the world, and how everything fits together. And so when I was in high school, I had the options of, you know, Spanish, French or Latin. And I was like, Oh my gosh, this is a second class where I can learn about history and pick up a new language. Let's go. And it turns out that a lot of the skills that you use in Latin and translation and the ambiguity, you know, it's simultaneously a very ambiguous language because nobody speaks it, but there are hard rules to it. It's a very similar brain space to electrical engineering. You know, there's, there's this hard V equals IR. You're either gonna blow yourself up or you're not. But there's a lot of ambiguity that comes into it, that I've as I've grown as I've learned as an engineer. It's a weirdly unique insight to have, and it's a fun party trick now too.
Aaron Moncur:Definitely, I was just gonna say, if you ran into one of the other 18 people in the world who speak Latin, could you hold up a conversation with them?
Matt Puchalski:Conversational. It's been quite a while, but I do really enjoy, you know, when I'm in a museum, less and less these days, you know, I sort of gravitate towards the Greek and Roman section to try to see if I can still, you know, read a transcription. It's always the pluperfect. That's the part. You know, a Rama rasa Ra, Mr. Cavickio, my Latin teacher, if he's listening, I'm sure he'll be thrilled.
Aaron Moncur:Very cool. All right. Well, you are a man of varied interests, one of which is robotics, and you're currently working on, or have already built three division systems for manufacturing. Tell us about that. What is bucket robotics? What are what is your team doing there?
Matt Puchalski:Yeah, so at bucket robotics, we are automating what seems to be. A very automated task, which is visual inspection. So our customers are folks who do injection molding and casting and machining, all areas of manufacturing that have been around for quite a while. But the fun part is, you know, very Silicon Valley of us. It's a it's an area rife for disruption on that quality inspection side, because when you really drill into it, something that is injection molded. Going back to our friend Ian, you know, he can tell you very often, the quality inspection the people who are really drilling into that surface fascia and saying, is this part good enough to sell or not? Very often that's done manually and in batches, which leads to one, very disgruntled employees and two, a lot of inefficiencies in your production, because you can't check every part, and so you may introduce you may see a defect, and then wind up having to bin, you know, an entire Gaylord, like hundreds of parts, because they're you you don't want to bother sorting through. So we solve that at bucket robotics with starting from simulation of the part. We take in the STEP file, and you can then build a model of what good and bad looks like without having to wait for for lots of components to go down the line,
Aaron Moncur:additional traditional vision systems, which I'm familiar with because we work with them quite often here at Pipeline, you might want to train it like You'll have what's called a gold standard part, and you'll train the vision system on this part. You'll get 30 perfect parts, 30 gold standard parts, and pass those through the vision system. And now the vision system starts to learn where the edges should be, where the features should be, all these things, but you're saying at bucket robotics, you have figured out a way to circumvent that potentially lengthy and expensive process, because you have to have actual parts by using the CAD model. Is that accurate?
Matt Puchalski:That's exactly correct? Yes. So the problem with the golden sample method that you know folks like key instance or some of the other huge players envision one that's predicated on waiting for those parts to be manufactured. And number two, one of the big problems that you'll see is 30 golden samples can very quickly turn into 300 or 3000 as soon as you enter into, you know, differences in colorways, or, you know, variability in how those are produced. And again, right? That's somebody who has to wait for those golden samples and then program your robot, and then give you that feedback loop. By starting from the STEP file, you have your perfect golden sample, which is, you know, you can't get anything better than, than at that engineering reference. And then from there we can take it. And you know, not just generate 30 or 300 or 3000 but 30,000 you know, 300,000 images of what good and bad looks like
Aaron Moncur:from the go that it's such a smart approach, right? I mean the cost and time associated with acquiring physical parts to use to train your vision system is not trivial, and it's such a great example of a disruption, because when you stop and like, take a step back and think about it, I mean, in order to procure a physical part, you kind of have to have a CAD model already. And the CAD model, like you said, is this perfect geometric visual representation of what the gold standard part should look like. So why not just use the CAD model, which is brilliant. How did you come up with this idea, like, what was the sequence of events that led to this aha moment? Yeah.
Matt Puchalski:So my professional engineering journey started when I was in college at Georgia Tech as a co op at Michelin at the research center in Greenville, South Carolina. And so at 19 years old, I was tasked with automation of quality inspection on tires. And so again, as you and most of your listeners will could probably guess an intern project and automation was, here's an Excel spreadsheet. You know, write a macro and make it a little bit more efficient, right? We still have this very manual process. But that really stuck with me. That really, really, you know, sat in my craw of like, Oh my gosh. Do you know how many people with flashlights exist double checking, which is great, you know, we want tire safety, like, that's, you know, should make you feel good about, like, buy Michelin. But just the monotony and the boredom and the horrible. You know, that was always there. And so I was very lucky to have worked on several different self driving car companies, and I've seen the bleeding edges of what computer vision can look like. So taking that, that initial idea, seeing the advancements, and then replaying every conversation I ever had with any of my bosses or any of my friends in manufacturing. Of you can't adopt cameras here because we're manufacturing something, a medical device, so we have to control our process. We can't have pictures leaking out. It's going to take too long because, you know, we need to keep the line going. So this needs to be perfect. There's too high noise and vibe of an environment like, Oh, you don't understand our process. It's, it's too shaky, right? And so you tick down the box of, gosh, all of these are problems that we have to solve at highway speed, and they're baby strollers and sunlight in self driving cars. So if we can solve it there, let's, let's try to bring that, you know, to a place that that immediately impacts people's lives who are producing these components that everybody uses every day.
Aaron Moncur:When you first had this idea, were there any, like, immediate roadblocks, you know, even thoughts that occurred to you. They're like, Oh no, it'll never work because of this, but let's try it anyway, and anything like that.
Matt Puchalski:Oh my gosh, yeah. So this is something that happens all the time in manufacturing and in engineering and in startups, right? Of oh, this is, this is probably a good idea, but who else is somebody? 10,000 people have had this idea before, right? And the answer is, yeah, there are lots of people who have had this idea. But what's that's deeply enriching, right? When we went through Y Combinator, one of the parts, one of the parts of the program, is that they really ask you to try to launch as early and often as possible, go out there and tell the world, because there's billions of people on this planet, and everybody is very, very smart. You are probably not the first person to have this idea. And so when we launched Hacker News, it was July of 2024, and the very first comment that came in was a plastics engineer saying, Hey, I have a master's in computer vision, and I do plastics. This is immediately helpful for me. I've thought about this for a long time. I'm great. I'm glad the free market solved it. The second comment was a PhD at ETH Zurich who was like, Hey, I've been trying to research this for years. This is awesome. Just cool. It's really fun to see the reception from the community,
Unknown:from motors and robotics to controls automation and precision prototyping. Simutek brings answer simulation tools and expert insight directly to your team. We help you accelerate design, reduce risk and eliminate costly rework, validate before you fabricate. Simutek group helps you simulate thermal stress, vibration, fluid flow, electronics, cooling, electromagnetic performance, you name it, whether you're an automotive, aerospace, energy, industrial automation or the next big thing. Simutek partners with innovators to deliver simulation driven results that move products forward, build smarter machines, launch stronger products, scale faster with simutex simulation support by your side, simutek group, your engineering partner in motion. Visit simutekroof.com and Let's power up your next design.
Aaron Moncur:Let's talk about we'll take a quick tangent here and talk about Y Combinator. For those of the our listeners who don't know what Y Combinator is, can you speak a little bit about that, and how did you land the opportunity to be a part of that?
Matt Puchalski:Yeah, so Y Combinator is the world's most successful startup accelerator. This is the the venture capital group that has helped develop and really launch companies like Reddit, Airbnb, astranis and boom supersonic, for and boosted board, you know, for our more hardware oriented listeners, and they're based out in San Francisco with this philosophy of make something people want. When we went through Y Combinator, they had a less than 1% acceptance rate from the volume of applicants. And they guide early stage startups through a program really, it's, it's like business school, but really condensed of kind. Out to San Francisco. Meet a few 100 other startups. You get paired with partners who have done this thing before, and we'll fund you to you know, really, let's see how we can make this as big as it can go. So in May of 2024 I applied with just an idea and strong opinions. And in June of 2024 I was in San Francisco with bucket robotics incorporated half million dollars of funding. And I was in, you know, I was lucky enough to meet folks like Brian Chesky from Airbnb, and, you know, Sam Altman from open AI and and also amazing batchmates who were on this journey as well building their their other early stage startups. These are folks I've grown to become, you know, deeply, close friends with
Aaron Moncur:incredible Wow, yeah, what an amazing opportunity. You mentioned that they were accepting less than 1% is that right?
Matt Puchalski:That is correct, and it's gone down since Yeah, keeps trending lower,
Aaron Moncur:yeah. So you had less than one in 100 chance of being accepted, and you were what? What do you think? All right, I have two questions. I actually have a lot of questions about Y Combinator, but we'll limit it to two and then get back to your technology. One is, what do you think you did that improved your chances? Obviously, you have a good idea. You had a good idea. But beyond that, what do you think you did that improved your chances of being accepted? And then the second question is, what are one or two things that you learned at Y Combinator that you feel like were the most impactful things that you learned?
Matt Puchalski:Yeah, so it's funny you say that the the idea actually is the thing that matters the least for Y Combinator. What they really love to fund is the founders. And how big of a problem space you are playing in, people who are, like, deeply militant about this is my idea. It needs to come through the world. Very often you learn that, like, Guess you're probably correct that idea should happen, but not in the form factor that you want. It's much better to say this is the problem that I want to go tackle, and I am the person to do it. So why me? So I was the integration and reliability guy for Jeepers. Okay, so I was at Argo AI, that was Ford and Volkswagens. So I've done Fords and Volkswagens and Ford again, and then Peter built trucks, and also Nikki a vehicle. So I've deployed on so many different form factors that that that's unique, Matt, but like, that's okay, that's whatever. But the problem space of manufacturing and having deep empathy, of, like, God, it's a really shitty job to be quality inspection person whose whole job is to say, is there a burn mark on this part? And to have that experience of, yeah, I've seen what good computer vision looks like. You know, Let's marry those two together. That's that was, I think, what led me to or bubbled my app, my application up, and then from there. It's just pure blind luck kind of thing, right place, right time. Nice.
Aaron Moncur:I also love that you say Volkswagen, you actually know how to pronounce, which I don't. It's not like I'm speaking from any level of superiority here. I just, I appreciate people pronouncing things correctly.
Matt Puchalski:No, of course, yeah. I spent a year of my life making the ID buzz autonomous with the the Argo and Volkswagen team. So it's really, really cool to see those out on in the world. Yeah, very cool.
Aaron Moncur:Okay, so second part of the question, What were a couple of the most impactful things that you learned at Y Combinator?
Matt Puchalski:Oh my gosh, yeah. There were so many different insights. I was so lucky, and they definitely seem obvious, but it's just very different to, you know, have somebody question you in front of you, right? One of the biggest things that Y Combinator really drilled into me is to always talk to customers, constantly be focused on what is the problem that you're solving. So even though we've, you know, I moved out from Pittsburgh, Pennsylvania to San Francisco for the program, I spent more than half of the batch flying to different cities around the country. And, you know, I would find a meeting on LinkedIn with some operations manager or quality manager. Wake up at five o'clock in the morning, local time, go to the Krispy Kreme, get two dozen donuts. Show up, you know, as early as possible, at 630 you know, when, when first shift starts, have the meeting, learn about the problems, and then ask, Hey, what are the three you know, what are your three tier ones that are immediately close to the this facility, and then it would go to the next one, and then go to the next one, and then get on the plane, and then come back here. And so very quickly, you know, you learn people take meetings before 9am local time. That's the best way. If you're a startup, don't ever schedule a meeting past 930 because, you know, that's when my day is completely blown apart. I'm sure that's that's true for everybody else. So if you can get a meeting at 730 that's the that's the sweet spot. And then the other insight from Y Combinator is to use your network as as much as possible. Again, going back to this, this is turning into a podcast about Ian. The reason, you know, the reason that I met him, is that he, if you Google injection molding San Francisco, he's the first one that comes up. And so I was just lucky enough, and now he's been, he's our first tester for everything, because his office is two blocks away from ours. And, you know, that's a nice and unfair advantage that we have of how many other people are deeply knowledgeable about the subject and are that close? It's yeah, it's been great.
Aaron Moncur:Yeah. Well, terrific. All right. So back to bucket robotics, when the team started actual development, what were some of the like, the early technical challenges that you encountered, that that you had to overcome?
Matt Puchalski:Yeah, so one of the big challenges is, we're playing in a space where there's one massive, player. It's the key ins Corporation, right? Their current market cap is like$93 billion which is a horrifying amount of money, right? And so one of the big challenges, that was a big forcing function for us is, we're a startup. We could light all of our money on fire by building a compute module or building a camera, and then nobody wants it, right? That's a form factor that nobody needs. And so that really helped us guide through our early stage. You know, one of our initial pitches was, we want to build a better Intel real sense. I don't know if you're familiar with this. It comes up a lot in sort of academic robotics. It's a really, really popular camera, and so really drilling into this, it became the philosophy of, why do people use this camera? The specs are kind of crappy. It's been around for a while, and the reason is, it's a, it's an easy software experience. So, you know, playing in the space where like are we hardware people? The answer is, yes, I deeply think of myself as a hardware person, but I become more software oriented as the company grows, because that is our, unique differentiator, being able to move quickly and actually build what people want. Yeah, nobody really cares about you know, how good of a quality imager it is. It's how easy is this to deploy? Will this solve my problem? Will this be reliable and robust? So, yeah, been a lot of fun
Aaron Moncur:speaking of key instance, who we love? We work with keyants a lot, and they're great. They have an excellent product. Have you ever seen the key and salesperson meme floating around the internet? You know?
Matt Puchalski:Oh yes, yes, yes. And and again, right? There's, there's at least two robotics companies and two blocks here. I'm surprised there's not a rap that you know has appeared organically
Aaron Moncur:in your bathroom. Yeah. All right. Well, we'll leave the meme at that. If you're interested, go Google key and salesperson meme on it's great. Okay, well, I'm going to take a very short break here and share with the listeners that the being an engineer podcast is brought to you by pipeline design and engineering, where we don't design pipelines, but we do help companies develop advanced manufacturing processes, automated machines and custom fixtures, complemented with product design and R D services. Learn more at Team pipeline.us. The podcast is also sponsored by the wave, an online platform of free tools, education and community for engineers learn more at the wave dot engineer. So we're speaking with Matt pacholsky today, and Matt, you've, you've alluded to this earlier in our conversation, but you've worked at a few places over the years. You've gained a lot of experience. Volkswagen Ford feel like I have to say that correctly myself now. And you've you've had experience with operations, managing Reliability Engineering. Are there any stories that you can share from your previous experience that that impart important leadership lessons do the product development expo or PDX is your chance to learn from subject matter experts providing practical hands on training for dozens of different engineering topics. Gdnt, advanced surface modeling, DFM, plating. And finishing techniques, programming robots, adhesive, dispensing, prototyping, tips and tricks and lots more. PDX happens October, 21 and 22nd in Phoenix, Arizona. Learn more at PD Expo. Dot engineer, that's p, d, e, x, p, O. Dot engineer,
Matt Puchalski:oh my gosh, yeah, that's a great question. So the beautiful part of I'm very lucky, the people that I've worked with across my career have been some of the nicest, kindest people on the planet, and to build robot cars that shoot lasers. And actually, you know, it's, it's such a it was a phenomenal time. And you know, the problem with that, or the growth opportunity in working in self driving, is that we live in a continuous time system. And you know, really mundane reasons can take down your fleet of self driving cars, and you, as the reliability lead have to report to the executives. Hey, the fleet in Miami is having a hard time as we come back from Christmas break, because over winter, a bunch of iguanas froze and they fell from the trees and landed on our LiDAR, and that's why they're all out of calibration. Is this, is this a real story? This is a real story. This is an actual thing that happens. Yes, again, like in a billion years, my tiny, little New Jersey brain would not think of this. No, no one would, yeah. Or, I mean, I had, I have a million of these. The it was always, it was always like that first week in January, with the cars fired back up, that the most insane things would happen. One One. Post Christmas break, we had a plague of laptops shutting down, and it was bouncing around between vehicles. And it started with one person, and then it grew to like five people. But we had 150 vehicles, and it was isolated to the operators who were sitting in the right side of, you know, the passenger seat, the laptops were shutting down mid operations. We're like, what? What does this mean? What is happening? And what we traced it to was people got Apple watches, and left handed Apple Watch users were triggering the laptop magnet that detected if the screen was opened or closed at just the right configuration. Oh, wow, and again in a billion years, you would never think about this, but that is a real issue that we had to triage and understand and go through the risks of what is happening. Is this safe to operate the fleet? How do you communicate to humans who are putting themselves into these vehicles every day? We understand your problems and we're working on them as internally, you're like, Oh my God. What the hell like, the sequence of events that needs to happen. Like, come on,
Aaron Moncur:how did you overcome that one? I mean, that's pretty unique. It's not that unique. A lot of people wear Apple watches on their left hands. Like, you can't tell people, well, you can't ride in our vehicle if you have a apple watch on your left hand. How do you get around that?
Matt Puchalski:So we did actually wind up banning the Apple watches. Yeah. They were Yeah. So, so it turns out that the Apple watches were affecting performance of the operations team because too many operators were like, you know, getting buzzes on their phone like they were looking at the watch instead of so it did wind up being a blessing to the ops managers. They were glad for that relief, but yeah, no, honestly, it's just building a really solid group of trust of operators being able to come to you and say, This is the problem that I am experiencing as I experience it, and then just going through that mitigation and remediation process of, okay, that's a really weird problem that you're having, but I can understand that you were in the middle of an intersection in South Beach or in Austin, Texas, or in Hamburg, Germany, you know, great. The difference of me being like, that's weird. Like, no, this is a panic situation for you. Let's, let's get you into safety and then understand what we need to dissect later, our skills that I developed with the robot cars, and it turns out, very useful to have as you're building a product for manufacturing where all of these crazy types of things happen every single day.
Aaron Moncur:Well, I'm sure you've had ample opportunity to use those leadership skills at bucket what? What are a couple of things, one or two things that you've learned about building a company that that no one told you about and you just didn't expect whatsoever.
Matt Puchalski:Oh my gosh, yeah, one of the things that I have. Really, really come to learn about building a company is to take care of your employees. And again, that sounds obvious, of like, you know, not to say that all of my bosses were, you know, Ebenezer Scrooge or whatever, but like, what that comes down to is it's just like having your own apartment for your first time, or, you know, becoming a manager at a larger company. Nobody cleans the bathrooms at the office unless, you know, you have the budget for that, and so I sweep the floors each night, kind of thing, and that is something that the team sees, and the team gives me feedback on, like, hey, we do have the budget. We can hire cleaners, but also it, you know, builds builds trust and builds a rapport that leads to a better product, and the team moving faster. So that's something that I never thought I would spend as much time thinking about. But I'm also very grateful for, you know, coming in and getting bucket sweatshirts and then getting the feedback of, Hey, these are a little long, but they're comfortable, and I like them like, you know, it's never perfect. You're never done.
Aaron Moncur:Yeah, I love your example of sweeping the floors at night, years ago, I don't even know how it started. I think one of our our team members just started doing it, and the rest of us saw him doing it and followed suit. But on Fridays, this one team member would just start emptying garbage cans all by himself, right? We have a cleaner that comes in every couple of weeks, but every couple of weeks is too infrequent to like, you know, get all the garbage, so oftentimes we just do it ourselves. And it started to become kind of like a ritual on Friday afternoons, like, well, the team, we'd all just, at some point, stand up and go around the office and gather all the trash bags out of the garbage and and take them out. And it was, you know, I had, there was an engineer who worked here at one point, and I remember him saying, not specific to the emptying the trash. I think it was like cleaning up in the warehouse or something. But he was like, you know, that's really not my job. And I'm I shouldn't. My time is better spent doing the engineering and not cleaning up. And I remember having a hard time with that mindset, very hard time. And so when it came to this, like taking out the trash ritual, I just I loved that everyone on the team was so willing to do this. It's not like, I got up one day, I was like, All right, guys, this is what's going to happen every Friday afternoon. You're all going to get up and take out the garbage. No, someone else started it, and myself and the rest of the team, we just kind of followed suit. And it was such a small thing, you know? But it was actually, I would say it was an important part of our building, our culture, where everyone chips in and, like, there's, you know, our Director of Engineering, Mike. When, when I hired him, Soon thereafter, I asked him if he was open to doing something I can't remember. And he said, Nothing is beneath me. Nothing is beneath me. And I was like, wow, I made the right hire. That's, that's gold, right there,
Matt Puchalski:completely it's, I completely agree. You know, the word that I love to use is empathy. I really use that a lot. I think it's really often underlooked in early stage startups. But that's the thing that helps you win in the long run. It's empathy for your customers. It's empathy for, hey, this is the same shared fridge that we're using. I'm going to, maybe not, you know, bring in something that's going to sit there. Maybe I won't microwave tuna, right? Like we're all, we're all human beings together.
Aaron Moncur:Yes, I have stories about microwaving fish as well, but we'll leave that for another day. All right. Well, Matt, looking forward, you know, over the next, I don't know, five years or something. What? What are you most excited about in the field of robotics and vision, manufacturing?
Matt Puchalski:Yeah, I am incredibly excited about the future of ruggedized compute and edge devices. This is something that, you know, when I look back and I started on self driving cars, the answer was, you know, raise a billion dollars and then take 200 million of that and get some custom compute that will arrive in two weeks or two years. And, you know, be thankful for it. And now there's, there's compute that you can buy from Nvidia, that will arrive from Amazon, right? That's, you know, in a form factor the size of a loaf of bread. And I see that, you know, becoming more and more of the default for robotics companies and allowing people to to develop faster and faster. There, and that gets me super excited, because then it becomes the, you know, then it's just the operational hellscape of like, oh gosh. Like, what, you know, what connector are we going to use, right? Like, what is our sensor modality? All the fun that comes with robots, of like, oh gosh, you know, so, so you're mounting hardware failed on you. What's the weird way that the whole system is going down? Right? But being able to start with that really ruggedized and robust computing is really going to enable so many different types of robot form factors that I can't wait to see what the next five years looks like.
Aaron Moncur:So a continuation to that question, what, what do you think interaction is going to look like in the future between humans and these robotic systems? We can talk specifically about vision, since that's your focus. But with, with, you know, the Advent and becoming more and more popular like augmented reality and virtual reality, and what does that look like? Do you think 510, years from now? How are humans interacting with these vision robotic systems?
Matt Puchalski:I am so incredibly excited. You know, it is my my advisor, through Y Combinator, is a woman named Diana who, who, you know, started a company, Escher reality, that was then became the backbone for Pokemon Go. So, you know, very lucky. Yeah, she's so I'm very thankful. And so it's funny because, you know, I think about this from my and operations experience all the way through what research has, and for vision, one thing that I'm really, really excited for is the sensor fused. It's camera and LIDAR. It's camera and time of flight sensing becoming much more available, much more ubiquitous, and that also opens you up to new and updated methods of, you know, really using computer vision algorithms. It's not just import open CV or pay a bunch of money to have a key instance, rap come and, you know, sit there and fine tune things. We're really in such an amazing explosion for really, really high quality vision systems. And there's a huge need for it, which, which is also something that you know, gets me excited from a growing business perspective.
Aaron Moncur:Yeah, vision systems are not cheap. I mean, we work with cognix and key instance pretty frequently, and it is not difficult to spend 10 or 15 grand on a vision system, and that's for kind of the smaller ones that get integrated into a machine, if you're buying one of their pre integrated, you know, desktop optics systems. I mean, you're like 8090, 100 grand,
Matt Puchalski:exactly. And again, you know your cost there is, will this line be an operation for enough time that I see an ROI, right? Or having to retool and repurpose oftentimes it's crap. I guess we have to buy another form factor, and that's another 80 grand, right? And that is a paradigm shift that I am seeing as we have more folks starting new manufacturing lines and thinking, Geez, that that isn't a sustainable model. I, you know, that's that's not the way that I want to do things. And that gets me excited.
Aaron Moncur:How about the this is a selfish question, the integration effort required for a bucket system, because going back to, you know, the behemoths out there, keans, condex, they've done some good things, and their software, depending on what you're doing, varying degrees of efficiency, I guess, and programming and integrating, but it still takes a while. Like, it's not like, you know, 20 minutes later, an hour later, you're up and running? It takes days or sometimes weeks, depending on what you're doing. What? What is the integration like for your systems?
Matt Puchalski:That's a fantastic question. We want it to be as easy to use as possible at every step of the way. So again, because we start from the STEP file, you know, it's a conscious decision. At every point of onboarding with bucket robotics, we start at the STEP file, because if you have dimensional CAD, you've got that right to the experience for how to actually integrate and use our synthetic data generator, right? That's that's what we're really doing. We're taking that step file, turning it into a data set of what good and bad parts look like. And bucket robotics recently won an award for our design for how to use AI and that human machine interaction. It's a deep obsession of mine, because I've also had to sit in those trainings before for you. Know, whether it's like National Instruments or whatever, right? It's like this. I'm not actually learning anything here. This is horrible. We're the, we're the anti vatic bucket robotics all the way through, you know, because we have such a very laser focus on simulation, we can simulate your camera parameters, right? It's not like and then buy some new weird lens. It's hey, let's enter in the parameters. We're going to make this as easy as possible. And here is a deeply tailored data set and detector for exactly that line. And so this is not, not two weeks. It's more like two hours and and again, it's all of the fun of, you know, it's two weeks on our end of making sure that we have the flights lined up to the newest facilities. That way we can actually show up with, you know, bucket robotic stickers and and donuts. And then we're also, you know, more geeks of like, oh gosh. What other problems do you have? Like, we have lots of friends in different, you know, startups like, how else can we help you? Because that's, that's sort of the way that that I see us winning.
Aaron Moncur:What's, what's the process like for a customer who is going to start using a bucket robotic system, you've got the like, the software, piece of it, right, the simulation, but eventually you're going to have to use hardware, a camera and a lens and actually start looking at your parts. Is the software hardware agnostic? Or are there like, recommended optic systems that you like using?
Matt Puchalski:Yeah, so we are hardware agnostic. Asterisks, right? So the thing that we have seen is there is always a chance that you need to change your you know lighting conditions, your you know your shutter rate, or things like that, depending on like, quite frankly, the laws of physics, right? But we have yet to see a customer that comes to bucket robotics that hasn't been sold a deeply expensive and overpowered camera when it's actually a software and inspection issue, it's Hey. So I have this crazy imager, but I have a part with lots of nooks and crannies in it, and so there's lots of shadows, and we're seeing lots of false positives. And the answer is, almost, you know, I yearn for the day that somebody has, like, hey. So we, our hardware constraint is, you know, an iphone four or something like that, right? It's never that. It's always like, here's a deeply expensive, you know, Swiss optics, crazy, but we can't make any sense of it. It's like, great, amazing. That's, you know, so I say that, and then all of my meetings tomorrow will definitely be with people.
Aaron Moncur:Okay, I've got one more bucket related question for you, and then maybe one more more general question, and then we'll wrap things up here. So the last bucket related question I have is, how can listeners know? How can people, engineers out there know, in general, if their product, their process, is a good fit for a bucket robotic system?
Matt Puchalski:That's a fantastic question. Our ideal customers are folks who are in the injection molding and casting. Injection Molding castings and stampings. Those are the start. That's our bread and butter. We really love working with folks who are on the finishings team. So if you have a stamped metal part, and there's lots of different ways that it can, you know, you can see a defect, and you never see it until after it's painted, like, oh, gosh, I just wasted all that time. Or even worse, it's painted and then installed on your sub assembly. If that's a pain point that you've had, if you're an injection molding shop that is, you know, a tier one to a large OEM, and you have a bunch of Class A surface, and you're paying people to make sure that you've got, you know, those nooks and crannies covered. We want to talk to you cool.
Aaron Moncur:I guess. The corollary to that is, what kinds of processes or parts? Is it not
Matt Puchalski:a good fit for fantastic, yeah. So let's go to, I'm sure, where the inbox is going to get filled up from. You know, if you're doing glass, if you were doing if you're doing glass, if you're doing like, Really 2d planer. So printed circuit boards, the free market has solved that. You know, if you're trying to see if my capacitor is is overheated or not, there's other products out there for you, but it. You're weirdly paying people to, you know, inspect the housing that that PCB is going into, because all other tools have failed you. Those are the folks that I want to talk to. It's the, it's my, my homies in, you know, commodities manufacturing,
Aaron Moncur:your homies in, commodities manufacturing. Greater sentence was never uttered. That's great. All right. Well, now one more here's the general question for you, and then we'll wrap things up here. What is one thing that you have done or observed to accelerate the speed of engineering?
Matt Puchalski:One thing that I have done that's really helped accelerate the speed of engineering, is to write as much and as often as possible. I think writing is such an amazing way to organize my own thoughts and come to a conclusion of, you know, this is the next step that I have for a problem that I am solving, and it's really, really fun. So you know, I won't say what my username is, but I'm a big fan of, you know, the injection molding subreddit and and, you know, even places like that, of just, oh, I can be helpful here. Let's actually, you know, writing is such a great way to advance your own understanding of a subject and to leave something for other folks so they can solve their own problems, whether now or in the future.
Aaron Moncur:That's a great answer. I don't think I've heard that one on this show. Anyway, writing as a tool to accelerate the speed of engineering, you're so right, right. No pun intended, but I've noticed that I've journaled off and on over the years, and recently I started again, thanks to my buddy Jake Kennington for getting me back on track there. But I think it's so amazing that as you start writing, whether it's personal journaling, reflection or writing about something technical, like a robotic vision system, you start to unlock these, these insights that seemingly come out of nowhere, right? I mean, they're coming from your brain, but, like, you never would have had them, had you not taken the time to sit down and, like, write something and really think through whatever it is you're going through. So writing is kind of it's kind of magic, isn't it?
Matt Puchalski:I completely agree. Yeah, I, I love it so much, and I don't do it often enough.
Aaron Moncur:Yeah. All right, well, you've got plenty of other things on on your plate, Matt. All right, I think we'll, we'll wrap things up there. So how can, how can people get in touch with you?
Matt Puchalski:Yeah, so you can find me on LinkedIn. I'm at Matt Puchalski. If you're interested in bucket robotics, we are at bucket dot bot. You can see a live demo, an interactive demo of how to generate a data set at app dot, bucket dot bot. And because I am an engineer myself, and the place that I go to learn things is YouTube, we also have a very deep YouTube channel explaining how, what the problem that we solve, how to work with bucket and yeah, I would love to engage with any of your listeners.
Aaron Moncur:Terrific, terrific. And that's at Matt Puchalski, P, U, C, H, A, l, s, k, i, which I definitely pronounced incorrectly before we started the show, so thank you for the correction, absolutely. All right, Matt, well, this has been great, super fun talking to you and learning about your company and the technology that you guys are developing. Congratulations on all the success that you've already had and all the success I'm sure is on its way for you and your team. And thank you for spending some time with us today and talking about it on the being an engineer podcast.
Matt Puchalski:Thanks for having me on Aaron. I appreciate you.
Aaron Moncur:I'm Aaron Moncur, founder of pipeline design and engineering. If you liked what you heard today, please share the episode to learn how your team can leverage our team's expertise developing advanced manufacturing processes, automated machines and custom fixtures, complemented with product design and R and D services. Visit us at Team pipeline.us. To join a vibrant community of engineers online. Visit the wave. Dot, engineer, thank you for listening. You.
