The Future of PCB Design Education: HaSofu Founder Kirsch Mackey

In this insightful episode of the Altium OnTrack Podcast, host Zach Peterson sits down with Kirsch Mackey, founder of HaSofu, to explore the evolving landscape of PCB design education. Kirsch reveals how his educational platform bridges the critical gap between academic theory and industry requirements, sharing his unique journey from physics enthusiast to electronics education innovator. 

Discover how HaSofu is revolutionizing hardware design education and learn about the potential and limitations of AI in PCB design workflows.

Watch the Episode:

More Resources:

• Connect with Kirsch
• The 8051 Microcontrollers & Embedded Systems
• STM32 Arm Programming for Embedded Systems
   

Transcript:

Zachariah Peterson

Hello everyone. Welcome to the Altium On Track podcast. I'm your host, Zach Peterson. Today we're talking with Kirsch Mackey, founder of Hasofu. You've probably seen Kirsch out there on LinkedIn and at some conferences. And if you're ever looking for educational material, he is definitely a guy to go and see. Kirsch, thank you so much for being here today.

Kirsch

⁓ Thank you for having me, Zach. Glad to be here.

Zachariah Peterson

Yeah, definitely. So we've talked a few times in the past and ⁓ I know that you've even done some work for Altium in terms of like some logs and some webinars. So it's really great to have you here today. But I think I don't know a lot about you and I think our audience doesn't know a lot about you. So if you could tell us your background, how you got into electronics. I always find that people who get into this industry have have some interesting stories. So I'm interested to hear yours.

Kirsch

Haha, right. Okay. Yes. So my background is that I'm ultimately Well, I'll start here when I was a kid. I didn't break apart electronics or anything like that Didn't have that like typical curiosity factor when it comes to tech in terms of breaking them open But I did enjoy like holding my first iPod touch my hand

And that inspired me to want to work for company like, specifically for companies like Apple, specifically Apple, honestly. So they were my dream company to work at as far as my career. And I had developed from a young age, from eight years old, started my first business. You know, we made like $146 the first week. And I was like, it was just crazy. So I loved science, physics, entrepreneurship, spirituality, all of these things together, just to understand more of how the universe works.

Fast forward to university, I start off in physics and figured I'd do electrical engineering because that was the closest thing to electromagnetic theory applied to get a job. But I figured I switched to EE because that'd be more practical to get hired. ⁓ My physics professor, told me, you don't strike me as an engineer. I was like, yeah, you're right. I'm a math and physics person. ⁓ And around

13 years old, I had got inspired by AI. It was a movie I saw with Haley Joel Osmond called AI Artificial Intelligence. And don't worry, this all ties in to where I am today. So I ended up getting into double E for my bachelor's, was gonna switch to computer science. This is don't flame me, audience. And then AI for my PhD, but I kind of stayed stuck in double E and made it work because it was still math and physics at the end of the day.

Fast forward, I did my master's and ⁓ did all my courses and my PhD, and now I'm working on finishing my dissertation for the PhD, combining AI and EE. In that time frame in academia, I'd gotten into industry and had to learn some additional skills, like power systems, power electronics, control systems, that eventually led me into PCB design as well as part of my expertise toolkit.

And in industry, I ended up landing my first role as a power systems engineer. Did that for a year and a half for big power grid stuff and analysis and studies with Python, then switched over to electrical systems design. So if you have a SunPower Equinox, I designed that, okay, at your home. And ⁓ eventually I switched over fully to PCB and hardware design. And now I'm here today and I teach. ⁓

⁓ electronics and hardware design at universities, my own business, and just do other things. So that's my background. ⁓

Zachariah Peterson

Very cool. So as somebody who, I guess, was going through academia not too long ago, we're similar, similar age, and looking at AI as a course of study, you've probably seen quite a bit of changes in how machine learning and AI are being implemented over that time period, right? I I think about

Kirsch

than two months ago, we're similar age. And looking at it in the eye.

Zachariah Peterson

seven or eight years ago, I was asked to review my first research paper on AI as a reviewer. And I feel like a lot of the stuff coming out around that time, seven, eight years ago, was like every third paper was like the 700th neural network for whatever application, right? was like everything was hyper-specific. And then fast forward to today, people were trying to do AGI with like one model to rule them all.

And I think the best approach is probably somewhere in between.

Kirsch

Yeah, same here. I've noticed that as well. it's similar to what they have in DeepSeq and certain models where you have multiple experts, where that's kind of where I was thinking and seeing things going, like you're talking about, having multiple experts in their own pockets, AI training each other to learn from each other. I mean, it makes sense. That's how humans do and build anything, right? So just copy and replicate what humans do. And now you've got

And Overlord that has many henchmen that makes it harder to take down. ⁓ But all that being said, yeah, I totally agree. ⁓ The AGI could maybe be the grand conductor, but having multiple experts in their expertise is a good strategy.

Zachariah Peterson

You know, I like, I like that you said that about AI, which is that it should, you know, be kind of like organized, kind of like humans are right. Because if you think about a company, right, you don't have like necessarily one person that knows everything. You have multiple people who are really good at, you know, a few things and they work within their domain and they work together towards a common goal. That makes sense. I think people should also have human level expectations of AI too.

because they definitely don't. have superhuman expectations of AI.

Kirsch

You

Absolutely. It's been, ⁓ I spoke with a number of companies and worked at one where the expectation was I want AI to do the entire PCB for me from start to finish, you know, as one of the examples, the entire electrical design. I'm like, yo, we have humans who are decades into this stuff and we don't even do that at full capacity, you know? So it's not, it's not realistic, especially

Because I think there are a miss sets of expectations versus what people know that AI can actually do. There needs to be more education in that. And then there's also fears that creep up. That people assume that AI is here to kind of take over. like, no, it's here to at best help us accelerate the things that we do that take up too much time that AI can do faster, like data collection.

There's a lot of stuff that could be learned from that and you point it out as specific, exactly.

Zachariah Peterson

Yeah. And speaking of human level expectations of AI, think if people did that, they wouldn't be surprised when AI fills in the blanks with information that is not totally correct or out of context or just made up. Because humans do that same kind of thing.

You asked me to give you something, so I'm giving you something, right?

Kirsch

was like...

That's a great point. Yeah, that's absolutely true, right? Yeah, humans do do that. ⁓ There is a post that I put on LinkedIn where I mentioned that this AI helped me accelerate my workflow from zero to like complete zero to place components on a PCB in like 15 minutes. But I intentionally didn't specify how or why. ⁓

People, especially the more seasoned engineers, they were rightfully extremely skeptical. They're like, hey, this is going to produce trash, foolishness. And that's exactly what you expect from someone who just got started in PCB layout design, just got started in electronics design, may not understand or know the deeper nuances, and doesn't have a guide or a conductor to help them navigate the paths where the more experienced person has been before. So yeah, people have to remember that.

It's modeling a lot of human behavior. Just remember that. So that would help.

Zachariah Peterson 

For sure, for sure. So going back to what you do professionally, you started a company called Haasafoo. I hope I'm saying that correct.

Kirsch 

Hmm.

It's real close. I like to call it hassofu, but hassofu is, yeah, yeah.

Zachariah Peterson 

Real close, okay.

Hasofu, Okay, hasofu

got it. Okay. I apologize ⁓ Tell us about about hasofu

Kirsch 

Right, Hasofu is ⁓ right now an educational organization where we teach people the skills they need to work in industry, starting with electrical and electronics engineering skills, specifically PCB and hardware design, okay, for right now. And that started as a freelance hardware company. So I founded that when I decided to do

freelance hardware design, along with another company that was offshore as well. Each of their designers already had like 10, 12 years of experience on me. And I came in about three or four years. But together, we were able to get some designs done. Some designs I could only work on myself due to ⁓ regional restrictions, limitations. So US-only boards, ⁓ only I could work on those. And that was great.

So great experience there. Eventually, I decided we'll change this into a learning institute and then eventually get back to doing services for companies and making our own products. How it got started, though, the name was from a project that I worked on and I presented at the Macmillan Design Innovation Competition back in 2017. And the product name was HESO, hardware software. Very creative.

right? And then I decided to bring that back for the domain and I wanted to get that but the domain was taken so I added FU which is a Japanese suffix for village or community so hasufu stands for hardware software community.

Zachariah Peterson 

I see, I see. And with this ⁓ material that you're creating, it sounds like you're really taking a holistic approach to get people ready for industry. It's not just PCBs, maybe it's not just mechanical. know, Hasso, the software part also plays a role. So I'm going to assume that's like embedded firmware. Maybe give us some of the topics that ⁓ you guys teach at Hassofood.

Kirsch 

with us on the tour of people. ⁓

Right, specifically.

So the topics we teach at Hasofu are printed circuit board design and then hardware design. What that means is PCB layout, electromagnetic theory, and applications. So signal integrity, EMI. Well, really, we start with EMC, EMI, signal integrity, design for manufacturing, high-speed digital design. All the fundamentals, all the grassroots

like even the intuitive understanding of the core equations, Ohm's law, Kirchhoff's voltage law, Kirchhoff's current law, those foundations. Then we teach the printed circuit board the application, also some professional software, particularly Altium Designer, for instance, and how to apply those skills to get industry level best practices, along with some IPC standards, IPC 2221. Then after that,

If you want to learn hardware design, we teach some electronics foundations. So your capacitors and ductors and discrete devices, component selection, peripheral and communication protocols, how those interact, and then circuit simulation. And so that's what we teach as far as hardware design. We also teach beyond that, like full stack hardware design. That includes debug, testing, embedded.

Code, C, Python. And then other than that, that's in the hardware ecosystem. Other than that, there's options for control systems. These are not publicly advertised beyond the full stack hardware design program. But there's control systems. That's where I started off in my expertise. Then it spans off to no power electronics. Some power systems, not yet. That's what I'm working on. And ⁓ that's what we've got. ⁓ and Python, of course, and some data cleaning. Yeah, of course.

Zachariah Peterson 

Yeah, I feel like you have to do Python these days. It's just one of those mandatory languages.

Kirsch 

And that's what we do.

Absolutely, absolutely. And then eventually AI, for sure. But that's, yeah.

Zachariah Peterson

So when you say full stack, you brought up test and debug. It sounds like this is hands-on education for test and measurement, that kind of thing. Maybe using scopes. Is that accurate?

Kirsch 

Yes.

Zachariah Peterson 

Okay, so students can get a, students can basically like get maybe a budget scope off the internet. They can get something off Amazon, maybe like a Digilent scope that Eric Bogatin has talked about and use that to follow along.

Kirsch 

Okay.

Absolutely. have my scope. Where's my scope? It is on the ground somewhere next to me, my four channel oscilloscope. And it's underneath my other electronics. But yeah, that's why my camera's not pointing all the way down. yeah. Yeah, we teach that. And students can literally do that and show you ⁓ what's the proper ways to set a scope, your probe impedance and what not, and how to read that.

Zachariah Peterson 

Okay. And then with, with firmware, I think one of the things that has driven so many people towards like popular microcontrollers like ESP32 and STM32 is most people have probably noticed if they've ever seen our one minute design reviews is ⁓ the fact that, you know, firmware courses really focus on those platforms essentially as like an example platform, right? So they can basically, you know, take their code that they developed and

they're developing it for like the dev board, then they throw it on the dev board, and they get to debug it, see how it works. Is that the approach that you're taking as well with the firmware development portion?

Kirsch 

Yes. Another part of the firmware development is teaching the 8051 in, yeah, starting with some assembly, quickly transitioning to C, but making sure to teach the hardware architecture, since there seems to be a, there tends to be a gap in between understanding how you get from like a chip's EEPROM to

Zachariah Peterson 

Okay.

Kirsch 

actually executing code. And so teaching that architecture, like memory addresses versus instructions and all that, and what's happening in the registers and the stacks, that seems to be aqueous for a lot of students. And so we teach how that works mentally. Once they get over that hurdle, then programming in C, Python, to make it do things becomes a lot easier. But yeah.

Zachariah Peterson 

So you've identified an interesting gap here. And I've actually felt the exact same way, frankly, ⁓ is that it feels like there's this gap between what you do in your electronics courses as an engineer. I mean, did physics when I was for my undergrad. But we also had to take basically those same courses. And you kind of get all the way up to memory and EPROMs, and then it stops. And then there's this big jump to, you

programming and it's like, well, what's all the stuff in between? How do people normally learn that? Do they have to go get a PhD in assembly language or VHDL or something and that's how they learn it? how do people normally learn that if they want to go design processors?

Kirsch 

Well, if they're trying to do stuff with FPGA, quite honestly, it feels like you have to get a PhD because my goodness, the VHDL very low. I mean, it's just crazy. Maybe because I personally don't like doing the carto maps and digital logic tables and everything. That's a whole other cheese ball game. But fortunately, you don't have to. If you have a course like called like microprocessor system design or a book by

Zachariah Peterson 

Yeah

Kirsch 

Mohammed, ⁓ it's a black book, and it's for the 8051. That is one of the best books you can get to learn and bridge that gap between that EPROM stuff and registers to actually what's happening in code from when the compiler starts doing its thing. That gap is completely put together in those 16 chapters in that book. He also has a version for the STM32 chip. ⁓

I don't know about the ESP32 yet, but he has one for the SEM32. I strongly recommend that book. That's all you need.

Zachariah Peterson 

Okay, after we're done filming, I'm gonna go and try and find that book, find a link where people can buy it, and we'll put it in the video description. So if you're watching on YouTube, check the description and see if I found it. No guarantees as of yet, but I'll do my best.

Kirsch 

Yeah.

Zachariah Peterson 

So, that's really cool that you've identified ⁓ this gap and I think trying to speak to it. And I would agree that maybe academia doesn't do a great job of filling in that blank. It probably depends what university you go to, what program you take, et cetera. I'm sure it's out there, but by and large, I think there is that gap. Are you picking topics at Hasufu?

that are based on these gaps in academia and then trying to fill those to get people ready to do practical things in industry? Is that how you pick all the different topics that you focus on?

Kirsch 

exactly how I'm picking the topics and it's because of the gaps that I had to deal with and face in my roadblocks that stopped me from being able to get hired. I wanted to be a control system engineer. I didn't know what exactly was missing ⁓ when it came to what I learned in the research lab in my classes.

between that and industry. I didn't learn that until I got into industry. And then the rejections made sense. I didn't know where the real gaps were between when I was learning PCB design outside of the academic curriculum, by the way, ⁓ and what industry needed for you to be a real starter. And I had to build that over time and eventually got it. Huge gaps. So out of sheer spite.

I made hassofu to give people specifically the bridge for what the companies are asking for in industry. You know, I had to get in industry and find out what they were really needing. That and what the university wasn't providing and still isn't and not yet. Hopefully they bridge that. That's why hassofu exists. It's to bridge the gap between university or academia and what industry needs.

Zachariah Peterson 

Ha ha ha!

So ⁓ what is stopping folks from industry, from going to universities and saying, hey, we want this? mean, I could imagine there are a few different things, right? Number one, they just don't have time. Number two, even if they did go to the university, there's no guarantee that the talent exists there or the knowledge exists there to teach it. What are some reasons for that?

Kirsch 

community.

Zachariah Peterson

Cause this is a persistent thing, right? I mean, you've identified something interesting here, which is like, you know, we talk about it in terms of like PCB design pretty often, but you're pointing out these other things that are also really critical for building actual products. And, know, you brought control systems engineering, you know, I don't do control systems, right? So I would not have even known that that is a pain point among people in industry.

Kirsch 

Yeah. Now, the fortunate thing about control system is that you do have a lot of that quote unquote theory that is actually literally very much applicable ⁓ to getting a control systems engineer role. Get some MATLAB, some signalling, do some stuff, and you're almost in. Build a real world project. And ⁓ but to build a real world project, you have to actually have real world skills that don't get taught in university. That's how I got into power electronics and PCB design. ⁓

Why are some of the reasons? whatever the reasons are for why I had the experience where I needed to learn how to use electrical harness design in other software when I was working in industry. And a $4,000 to $6,000 teaching course from

One of their customers, one of our customers who sells us the software, you use their software, wasn't approved by the manager, right? A lot of education doesn't get approved by the manager, even for their own employees who want to level up, despite the fact that the senior engineers don't have time to teach them. Senior engineers are too busy. Maybe I guess they're understaffed and they don't want to spend any money or resources, don't have the budget to.

teach even their own. So if they're not even going to do that within industry as much, then I think it's that they need a compelling business case for a return on investment to go out in university and bother doing it. It seems like it'd be much easier to just find people who are already ready to go and hire them and solve the problems now. ⁓ For companies that have

like huge budgets and are intentional about educational programs, two-year programs. mean, those do exist. That's when I worked at ERCOT and everything. They do have programs that are more geared toward students. So they exist. It's just that it's like you have to get to the biggest level as a company, like General Motors, General Electric, Apple, ERCOT, to even be able to have the bandwidth and funds for those programs, it seems like. That's what I've noticed.

I don't know why. Other than that, it seems like industries could just literally go into universities, teach these students, instead of just doing pitch campaigns to just work at their company. I don't know.

Zachariah Peterson 

Yeah, the pitch campaign seems self-defeating too, because they'll do the pitch campaigns, they'll get a hundred applications, let's say, and then, you know, if the skill isn't there, they're going to reject them all. It's like, well, why did we even do this? It's just a big waste of time for us.

Kirsch 

Yeah, it's a huge pain point, know, a huge set of pain points. I've been through that. even lost motivation for career fears because of another set of skills that I won't even get into, which is marketing your resume and, you know, selling yourself, pitching and all that stuff. That's a whole other set of things, but that's just, it would, it would definitely, it definitely seems like a waste of everybody's time on both sides. And I don't know why it keeps happening.

Zachariah Peterson 

Sure, sure. Now, I think the other side of that too is, sure, a company can go into university and talk to the department head of let's say the E-department and ask them, hey, we would love it if you would teach this. I think it carries a lot more weight if General Motors or Intel walks into that office and says it versus ABC Corp who has 30 employees.

Kirsch 

Great point. Yeah. So that opens up another thing. Yeah, universities, are also, you have your for-profit universities and your not-for-profit academic. So it depends on the motivators for the university as well. know a number of books get, like straight up has been told to me by a professor, a number of the books they use are chosen from

based on incentives and partnerships, not necessarily the best book for that particular curriculum. ⁓ Yeah, I used to skip.

Zachariah Peterson 

Interesting.

Kind of concerning in a way, if you think about it.

Kirsch 

Very much so, to the point where I used to skip the classes of one of my courses, but I enjoyed the topic so much, I skipped the class because I was not, the book didn't make any sense. I got another book that cost a third the price and did perform better than most of the students in the class because of that other book choice. So there could be that, there could be like,

To speak to your point, actually that's a great point. I noticed that when there was ever a change, it would be a big corporation like Texas Instruments, for instance, putting oscilloscopes in a whole lab and then putting their whole branding on the lab. And that's when change happened. So that's a great point you bring up. be commercial brand kind of things.\

Zachariah Peterson 

Interesting. Okay. Well, I don't know, maybe we'll see the Altium lab at universities at some point and actually drive it the direction we want to drive it. And that probably explains why when I was taking a graduate thermodynamics class, they were using the exact same book as the undergraduate thermodynamics class I had been in. And I'm like, I'm learning nothing right now. And I feel like this is a big waste of my time and money.

Kirsch

Hello.

Yes, yes, absolutely.

Yeah, that's been the case for a lot of my classes. I'd say at least 33 % of them. ⁓ However, also to give a little bit of a note, he also did mention that they are responsible for choosing books that do cover a wide enough range of understanding. So a lot of times, there aren't too many textbooks that go into sufficient depth and range.

to cover what they need to cover for the students. Because universities are also tasked with making sure we are exposed to enough of a topic within one semester for that class. So it's a bit of a balancing act as well. But in too many cases, they just could have chosen a better book, honestly.

Zachariah Peterson 

Yeah, this gets into another problem. Well, maybe not a problem, but I think just something that happens with engineering education is a lot of times topics are covered, I guess you could say, ⁓ a mile wide and an inch deep, right? They try and at least ensure that you know the vocabulary and what is this thing at a basic level. And then it's like, okay, on to the next thing.

Kirsch 

That's exactly right. Yeah.

Zachariah Peterson 

Okay, okay. So one thing, you you talked about artificial intelligence earlier, and I'd like to get back to that momentarily, because you also mentioned that this is something that you're looking to get into with Hasafu. And I'm wondering, what would students be exposed to if they were to engage with Hasafu on an artificial intelligence course? Is this like,

LLM development is this instantiating LLMs and hardware, designing hardware to support that level of compute. Give us kind of the overview.

Kirsch 

Whatever is most important for industry right now. So one thing would be, if they're in hardware design, understanding PCIe and PCB layout optimized for running TPUs and GPUs in particular kind of hardware, that fast throughput, so high speed digital design. For people who are specifically trying to get into AI, absolutely, it's interesting you mentioned that. Building an LLM.

Well, not building one from scratch not yet, but training LLMs and deploying them for your own local applications and then one that you can rub run on an AWS or you know any kind of web server to just Start making money, you know because Yeah pulling quen 2.5 O llama all those that's what they would encounter something that would help them produce something that would help them make products

which is what companies are looking for, products and services, things that make money. Another thing they'd encounter is the mathematics. If they're trying to get deeper, the mathematics behind the different models in AI, machine learning, linear regression, KNN, convolutional neural networks, that's a big one as for vision, for visualization and whatnot. Data cleaning, that'd be like 80%, honestly, of things.

Yeah.

Zachariah Peterson 

Yeah, yeah, I can't believe how many people fail to grasp that like you get an accurate LLM by picking the data, not by just prompting.

Kirsch

Exactly. So that's what they'd be exposed to at first. But also math kind of early. ⁓ Even the idea of the probability, right? The deterministic and probabilistic kind of more. OK, so yeah, I'm trying not to get into the quantum things. But anyway, the more probabilistic parts of that's not probably in stochastic processes and linear algebra.

of AI, ⁓ but in a more intuitive way so that they could understand and gauge the benefits and drawbacks of the different models, including GPT as well.

Zachariah Peterson 

I think those are important topics, especially at the math level, because if you're going to do any kind of instantiation of an LLM in an embedded device, right? Whether, like, I don't care if it's a microcontroller, FPGA, whatever, right? If you're going to do that kind of instantiation, you really have to know something about the math and not just, you know, pruning and those types of processes on like a neural network. You have to know something about the math because that's how you're going to

even pick the hardware platform that you're going to target and then try and optimize how you actually run the processing in order to produce inference results in any reasonable amount of time and without this boatload of memory on the device.

Kirsch 

Absolutely. And this gets into the edge computing. So ⁓ we don't have a curriculum for edge computing yet. ⁓ But that, you'd think, for us teaching hardware. But that's on the way, maybe 2026 or something. ⁓ But we're trying to bridge that gap in an elegant way. So we're hitting.

Zachariah Peterson 

Yeah, yeah.

Kirsch 

Both of the hot topics ⁓ that are paying right now. It's not that edge computing is not paying, but it's just that it's in its developmental stages still, I would say, in terms of ready to go super fast commercial popularity. Let me say that, popularity. Popularity is your driver.

Zachariah Peterson 

I would agree. mean, it's definitely, it feels like the Google search volume for edge computing has gone way downhill ever since our artificial intelligence became popular.

Kirsch 

Yeah.

⁓ that's interesting. I haven't looked at the stats, but I about, know, yeah, I would be, I'm afraid to look at them and see how terrible this disparity is.

Zachariah Peterson 

Well, you know, I mean, I have to look every so often, right? But I haven't looked either, but really I just look at what people are talking about in some of the big publications, right? You just look at some of the big publications and like, how do they pick what they put into a title, right? They're picking it based on search volume because they want people to land on the magazine and read something. Obviously that's why they even exist. that, you know, you can really kind of gauge what's popular just by looking at some of those big publications.

Kirsch 

Hmm

Zachariah Peterson 

in my opinion.

Kirsch 

For sure, that's a great point. That's a great point. I have to take a look and see. mean, I do have a project on edge computing where you would, I haven't done the project yet. This is just a proposal for one of the journals where you'd have edge computed devices, let's call them that embedded devices that you put on the grid to do that pre-processing on the board, right, of phaser measurement unit.

The measurements on the phasor measurement units or PM use take 60 measurements per second lots of data terabytes gigabytes, know, it's big data and Then process that pre process that before it even gets into your internal ⁓ Measurement units and systems well in turn into your internal let's say Processors so to speak and data collections for the grid. We want to do a lot of that cleaning on the boards first

before we get that into the central power grid storage collection things. I'll just call it that. ⁓ So that's where edge computing would really come in and be extremely helpful. But yeah, I haven't done the project yet.

Zachariah Peterson 

Yeah, that makes sense also. mean, one of the things that had been talked about with edge computing for a while was like doing embedded AI. Like we're just going to like replicate what happens on the cloud on the edge. And I think that's really something that's more appropriate for like specialized use cases, including like military where they really may have internet denial going on. So they don't have a choice. They have to do everything at the edge. But what you're talking about, pre-processing data at the edge.

And then you just give it back to your central repository and you're to run AI on it, then that can happen in the cloud. You know, that's a really good balance between trying to, you know, do literally everything on this tiny device that's supposed to be deployed at the edge. Cause I think at some point that becomes unfeasible.

Kirsch 

Yeah, I'm like, when my professor proposed the product, was like, well, how much are we processing on these boards right now? And where's the budget going to come from to make these? Like, what? So yeah.

Zachariah Peterson 

You

Yeah.

by the way, it has to run on a coin cell battery and you know. ⁓

Kirsch 

And harvest solar energy, know, and work at night like, like what? So, but yeah, he and I ⁓ quickly brainstormed that and like, yeah, we just do part of the processing on the board and the rest in the central system, just like we, yeah, relieve some of the load. Yeah.

Zachariah Peterson 

Yeah, yeah.

For sure. Well, I hope folks that are listening right now will check out Hasafu and will learn or find a topic that really interests them and get some of those skills and level up. So what are some of the other future plans for Hasafu?

Kirsch 

The future plans for Hassofu is to release a course on not just PCB or hardware design, but power electronics specifically. So switch mode power converters starting with an AC DC. That's becoming, it's not that it was never there, but it's becoming more prevalent. I noticed people who want to get into the automotive industry and motor drive design. It's kind of like, do I even, like, how do I even start?

kind of thing. Example projects for people who are wondering. You could do example projects. So there is that. Power electronics course. then after that, do have this may be a lot, but 40 courses that we're working on that would hit each level of expertise or level that someone's a learner at. So a beginner who has no idea.

about the hardware design ecosystem. And then someone who's already designed some circuit boards before, and they have some idea, but they don't know how to break an industry. And then people who are in industry but who are trying to get to junior or mid-senior level, courses for them. they all teach the same foundations, electronics, part selection, all that stuff, but at the appropriate levels. And then the last level would be your industry professional who can do everything in the world.

That would be at the level of designing your cell phone, gaming system, making your own commercial products, that kind of thing. So that's a course plan. We also have product design courses. We're coming up with a toothbrush course pretty soon in May. It'll be in Altium. Everybody's asking about Altium. Ask people. It'll be in Altium. And then after that...

Zachariah Peterson 

Yeah. ⁓

Kirsch 

In 2026, we'd have a boot camp for AI engineering specifically. That's starting to be a bit more popular. I don't know about popular, but a bit more, ⁓ and it makes sense, a bit more targeted than, say, just data science or data engineer. AI engineering specifically is what a lot of companies are looking for.

Zachariah Peterson 

Yeah, well, I think a lot of people thought that the AI bubble would have burst by now, but ⁓ I think it's definitely not going away. I think it's here to stay and it's only going to grow. And I think it's very smart that you're targeting that kind of content for folks who want to get into that side of the industry.

Kirsch 

Thank you. And speaking of which, I forgot to mention a huge thing, the AI integrated hardware design workflow. I completely missed my mind. So using AI software that accelerates the path toward creating hardware in addition to your standard software, your standard ECAD software. So yeah.

Zachariah Peterson 

Okay.

Absolutely. Think that's really cool too, because I think right now it's kind of the Wild West. And there's a lot of great startups that are trying to develop solutions ⁓ that can help accelerate certain tasks. And I'm really pleased to see that they've taken kind of a task-focused approach to build these tools. I think that makes perfect sense. Because this idea that you're going to just

prompt into a chatbot like, build me the iPhone 24. I think that's just a total fantasy.

Kirsch 

It completely is. It's like, it's like, I don't know the complete progress on self-driving vehicles, but I always think of that as an analogy to where I'm like, hey, unless I can trust an AI to drive me around for 10 years straight without an accident, it's not fully self-driving to me. I don't care what anybody says. Cause as a human, know humans who've been able to do that. Like that's the minimum.

I needed to be at. So yeah, before I could trust that, or even with programming, software development, I use AI to produce programmable, like program back end dashboards and everything. And software engineers and myself included, we've experienced it where they will rewrite certain things that it's not supposed to. So even in software development, there are bottlenecks, there are issues. Harvard is out development. Good luck. That's not going to happen. It's not realistic.

Zachariah Peterson 

Yeah.

Yeah. And I don't think it actually levels the playing field between hardware people and non-hardware people, even if that were possible to just prompt a chat bot and it gives you a ⁓ printed circuit board, because at some point you have to actually speak to the bot as a hardware engineer, just like you have to speak to a bot as a software engineer if you want to get useful working results out of a bot in terms of software.

Kirsch 

Absolutely. ⁓ The important thing about AI is that it's going to make the subject matter experts more efficient, get more opportunities, and earn more money. And the ones who aren't already experts, can learn some things. But it's

going to give them, if they were to use AI, let's say they were to use AI to design hardware, there would be so much to figure out debug. You still need to understand and learn the foundations, the physics, the best practices. The gap is wider. Those who have the expertise will move faster. Those who don't, won't. ⁓ And yeah.

Zachariah Peterson 

So what you've pointed out here, and I know we're getting up on time a little bit, but I want to bring this up because this is really interesting. I've told many people that when you have a disruptive technology like AI, whether it's hardware or software, or even just rewind 100 years to mechanized farming, those who have the new technology can experience a bifurcation where the people who have it know how to use it.

They go on the upward slope after that bifurcation in terms of opportunity and probably financial success as well. And then the folks who who fail to utilize it properly, they go on the downward slope after that bifurcation where they then lose opportunity and probably also lack of financial success. Do you think that's fair in terms of hardware engineering?

Kirsch 

I would say so. I would say so. On the flip side as well, AI that I've seen has helped to democratize or

Productivity.

Zachariah Peterson 

Well, sure, sure. I think it can

democratize productivity, but it doesn't guarantee anything, right? You still have to make the best use of it in order to the success and the productivity increase and ultimately the financial gain that it supposedly promises.

Kirsch 

Yes, yes. I was going to have a go in the direction of there are tools out there that do have some AI to do some of heavy work to, you know, the tedious stuff to get something out if you're a beginner or hobbyist or whatever in terms of hardware product. So you can get fast, you can get to prototype faster. And it does a pretty solid job. ⁓ At the end of the day, though, you have to consult an expert, you who knows what they're doing.

So at the end of the day, it accelerates them toward, it accelerates the not as experienced people or non-experienced people toward getting something done. But the gap is widened, like you'd mentioned, where, yeah, no, they can't accelerate faster. It's at the limit. Their cap, their ceiling for the use of that is how much they understand about it.

That's the perpetual ceiling. AI only accelerates and helps you when you already know how to do. And if it produces something you don't know, you have to consult an expert anyway. So it's still at the limit of whatever human is using it. So I guess your point is that it's making it more obvious. Yes, there we go.

Zachariah Peterson 

Yeah.

Right.

Who's the expert and who isn't? Yeah. Okay. That's fair. That's fair. I'm, I'm, I'm still waiting to start getting consultation requests for, you know, like we use chat GPT to help us design this and it doesn't work. We don't know what we're doing. And that speaks exactly to what you're talking about.

Kirsch 

Thank. Yes, and also to what you talked about how they use it too, like if they have a misunderstanding of how to use it for those kinds of things, then ⁓ that will definitely lead to that. I find it interesting then that certain SaaS companies are, I don't know if they've noticed that early on, probably have, which is maybe why they're targeting like the tedious stuff and not the

⁓ More agentic stuff because they're like no this is gonna this is gonna be this is just gonna be a mess if ⁓ If people use this incorrectly, so yeah the chat GPT Persons if you're using it for a piece of be loud and also be very careful. There's one additional piece of warning This is what scared me when I use chat GPT To help me explain why you would

Layout a PCB based on a particular wavelength or whatever transmission line effects. It gave me the right answer, but the inverse reasoning. That's very dangerous. If I didn't already know it, I wouldn't have been able to catch the reasoning. It's all about the reasoning. Hardware design and engineering follow a

Zachariah Peterson 

Hmm.

Kirsch 

deterministic set of rules, best practices, and real world physics that does not change. ⁓ You don't want to be using LLMs ⁓ in a deterministic based process, especially with math and physics. So please consult a subject matter expert. You can reach ⁓ seemingly correct conclusions and even correct conclusions. But with the wrong reasoning, your judge, I mean, your bridge will fall apart. Your board will fall apart.

Zachariah Peterson 

Yeah.

Yeah. And that correct, correct versus incorrect reasoning. If you take the incorrect reasoning from the output and commit it to memory and then run with it everywhere else, now you're just making a mistake over and over and over again and not realizing it.

Kirsch

Thank you.

And exactly, this is exactly why companies, when they're interviewing for hardware design roles, they ask about why and how you choose your PCB stack up, why you choose smaller package capacitors. They want to know how much do you know about the impedance, you know, ⁓ the inductance levels of capacitor leads and all that stuff, the signal integrity. They want to know your reasoning. And they hardly ever ask about the complexities of your boards.

They want to know, do you have the right reasoning? Because with the right reasoning, you'll always draw the right conclusions and the best practices. ⁓ And another analogy that comes to mind is when professors and teachers want to see your reasoning behind a math solution instead of just a right or close evaluated number. They want to know how you got there because you could have gotten the right conclusion with the wrong reasoning.

Great perfect recipe to go into field and break something. So yeah.

Zachariah Peterson 

Well, I think we're going to leave it there because that's really great advice. And folks can come to Hasofu. I said it right, correct? Hasofu? OK, good. Yeah, folks can come to Hasofu to learn all this stuff and learn the correct reasoning for what we do in hardware design. Kirsh, I want to thank you so much for being here. It's been too long not having you on. So I'm great to have you here today. And I'd love to have you come back in the future.

Kirsch 

Yeah, yeah.

Well, I appreciate it is never too late to be on the on track podcast.

Zachariah Peterson 

Hey,

there we go. We're going to, I'm going to start using that. Thank you so much. To everyone that's out there listening, we've been talking with Kirsh Mackey, founder of Hasofu. You can, ⁓ I can't believe I just messed up the outro. Let's try that again. Okay. Joel cut me back in. To everyone that's out there listening, we've been thought, one more time. All right. To everyone that's out there listening, we've been talking with Kirsh Mackey, founder of Hasofu.

Kirsch 

You

Zachariah Peterson 

If you're watching on YouTube, make sure to hit the like button, hit the subscribe button. You'll be able to keep up with all of our podcast episodes and tutorials as they come out. And last but not least, don't stop learning, stay on track, and we'll see you next time. Thanks, everybody.