Transcript Judy Warner (00:01.157) Hi Colin, thanks so much for joining us on the podcast today. It was very nice to meet you at PCBWES. I'm excited to share you with our audience. Collin Stoner (00:10.382) Thanks for having me, Judy, I appreciate it. Judy Warner (00:12.911) Well, as I mentioned, I was at PCB West recently and was kind of overwhelmed with how many startups and entrepreneurs were there creating different solutions for design engineers, leveraging AI and LLMs. And so I wanted to make this whole series, which you are part of now. So why don't you start out by introducing yourself, tell us a little bit about your background, and then give us a brief overview of Zeno.ai. Collin Stoner (00:43.15) Sure. So my name is Colin Stodder. I co-founded Znode.ai with my friend Brandon. I'm an electrical engineer. I have a degree in electrical engineering and another in computer engineering from Michigan Technological University. And I've been running a consulting firm for almost the last 15 years. And since I started serializing my designs in 2016, I've done about 250 or so different circuit boards for all kinds of customers, everything from startups to NASA and everything in So I started Znode with the idea to help solve a lot of the problems that I experienced when I was doing the designs. I experienced a lot and I'm sure a lot of our listeners have too. You're working through a circuit board and it's just monotonous clicking over and over or scrolling endlessly through data sheets or manually copying pin assignments in when they're doing the symbol. And it's a lot of work that seems like there should be a better way to do this. But every time I think about it, as I'm doing the design, it's like, well, not really because this is special or this is a little different and it'd be really hard to handle with automation. And that was true until about 2022 when all of the new crop of AI tools came out. It's like, hey, actually, maybe we can automate some of these things that take me forever and are necessary but don't add a whole lot of value to the design. And so that's what, you know. made me want to start a company for electrical engineering. Judy Warner (02:25.433) So when the 2022 season hit, what were the things you were seeing that became perhaps some enablers to solve those more nuanced pieces of the design and what are the overall problems you're trying to solve with Znode? Collin Stoner (02:43.532) Yeah, so I think the new generation of really chat GPT made it obvious that you could do a lot of these things. Now there's a lot of problems with things like chat GPT, which I'm sure we'll talk about, but. What inspired me was the way that it could handle, I'll call them fuzzy inputs. And that's a lot of the problems that I saw with automating these things when you're doing a board design or an electrical design or really a lot of other kinds of designs. There's just sometimes you have to do a lot of repetitive work, but it's not completely repetitive where you can just copy it into a spreadsheet and hit a few buttons or write a little bit code and do it all. It requires a little bit of thought. or looking through stuff, you have to put a little bit of thought into what you're looking at. Not a lot, but enough that you have to do it. You can't write a program to do it. The new AI tools make it possible where they can do that, and they can do it at scale, which is what gives us the ability to make a lot of really interesting tools. So. what we're trying to do at Znode is make use of that so I can spend my time as an engineer doing things that I actually like doing as a electrical engineer. Designing the schematic is one of my favorite things to do, designing up a new design. But going out and finding just the right op amp that has just the right specs that I need can take a little bit of time. Or. Collin Stoner (04:19.326) know, finding the FPGA that I need. That can take a little bit of time too, quite a bit more than an op amp. And finding all those options is kind of the first thing that we're working on with Znode is using an AI to go through the data sheets to get me the information that I need to make my decision a lot faster and let me see the options as I want to see them, not how they're sorted by specs that I'm interested in, but also things that I can't see on the top page of DigiKey. You know, how do I program this microcontroller? What are the tools that I need to do that? have this tiny little feature in the ADC that I need so that I can reference, you you use this other voltage reference. You know, lots of little details. I want to see all that stuff. And I'd like to not have to read 20 data sheets before I find the right thing to do that. So that's what we're building now. Judy Warner (05:14.055) Why not? It's so fun. I'm sure. So you, you mentioned chat GPT and it's early days and it could be messy. how exactly, to me, what you're saying is, you know, everybody's using this co-pilot or AI assistance, right? But again, it's early days and it's messy. So how exactly are you? leaning into the way you leverage AI or LLMs that fit well and are accurate for design engineers. Collin Stoner (05:50.454) Yeah, that's a great question because I think we have a different philosophy on how we want to use this. I'm an engineer. think most of the listeners are too. I don't want to not do engineering. That's really what I like doing. I want tools that make me better and faster. So we don't want an LLM to come in and try and design our circuit for us. And I don't want to make an LLM that attempts to do that. What I do want to use an LLM to do is dig into the data sheet when I ask a question and find me the pertinent things that might be useful to me as I'm doing my design. What's the I2C address of this chip? Well, there's four or five sections that talk about it, and for some reason, no one ever just says the I2C address is this. It's always in four charts with six different options and... I'm going through a schematic and I have 17 I2C parts and I'm trying to build out a table of them. That's not super interesting electrical engineering work. It's kind of busy work. I need to make sure that I'm right. So I still want to read the sections and make sure, but I want to use an AI to help me find those quick or find the little details about a part as I'm searching for it. So I want the AI to do the heavy lifting of Judy Warner (06:55.472) Yes. Collin Stoner (07:12.63) searching for stuff across lots of data sheets or finding me things that are out in the world in the ecosystem of stuff that it's been trained on that might be useful in my current situation, whatever that might be as I'm doing the design. So we want to make tools that make you faster and better as an engineer, give you the right information at the right time to make the decisions that you're trying to make in a more informed way. aren't really smarter than you, and they're not going to be for a long time, but they are faster. Judy Warner (07:47.033) Yes, exactly. think that's a very important thing to say. And again, in the spirit of being an assistant or parameterizing it in a way where just all the good stuff you care about comes to the surface and then you get to choose and use your engineering judgment intuition. Collin Stoner (08:07.09) Yeah, use it's speed because that's what it's good at. It can search through and read a lot of stuff a lot faster than you can and give you the cliff notes so you know where to start looking for what you want rather than having to go find it yourself. looking through data sheets is not the whole job. And that's just one example of where it can be faster for you. But we want to apply it in lots of places where having that ability to read quickly at scale Judy Warner (08:13.476) Mm-hmm. Judy Warner (08:17.797) Mm-hmm. Judy Warner (08:29.735) Mm-hmm. Collin Stoner (08:36.957) or apply lots of knowledge at scale to find you the right spots to look for. That's what we're trying to leverage. Judy Warner (08:44.719) Now, just in general, I'm sure we don't have the time to dig into the details here, but there's some other companies that are saying they're trying to do the same thing. So, you you've been engineering for a while across a multitude of designs. So when you think about this problem, how to solve it, what's different versus maybe some of your competitors or other startups that are trying to tackle the same problem. Collin Stoner (09:12.172) Yeah, so I think there's a lot of folks that are trying to start at the top and do the big sexy, let's generate a design from scratch or from your instructions and be the end of what you have to do for design. But as I just said, I don't want to do that. I want to start from the bottom up and use AI where it's useful today, not where it might be in a couple of years or when some magical AI comes out. I want tools that help me now. And so we're starting at the bottom and applying it where it's useful today. So the first one of those is searching for parts. This is a fairly straightforward thing to do as an electrical engineer, Judy Warner (09:36.551) Mm. Judy Warner (10:00.507) Mm-hmm. Collin Stoner (10:01.228) We've all been on DigiKey looking for a part that meets whatever needs we have for whatever we're working on that day. And a lot of us have gotten really good at finding the parts that we need. But I feel that a lot of times there's something missing there. And it would be great to be able to just say what it is that I want and have it look through not only the parts, the specs and the stuff that you might find on the front page of DigiKey. but also the stuff that's in the data sheet or in some of the technical manuals or in some of the errata materials about that chip. What are the things that are unsaid in those manuals that might be somewhere else that's a little hard to find that I'm going to find out when I've built this board and then it's like, ooh, I did this thing and we have to do a respin. So I want to see those things up front and I've got 20 chips to look through. Judy Warner (10:41.639) Mmm. Judy Warner (10:50.683) Respin. Collin Stoner (10:58.254) know, give me tools that help me do that faster. And so that's where we've started. So we made a tool that lets you search for parts, but doing it in natural language. I mean... Judy Warner (11:03.654) Mm-hmm. Judy Warner (11:09.607) So in that natural language, give us just a quick example if you could like say you're looking for an op amp or ADC, know, some of those nuanced queries you might be able to make in zeno.ai. Collin Stoner (11:23.148) Yeah, so I had an application. We were working on a UVC LED driver a couple of years ago, naturally, during the pandemic. And the LEDs are such that you have to drive them with constant current. So they're not like visible LEDs where you can PWM them and everything's fine. So I needed a constant current output LED driver. The specs on DigiKey don't necessarily, and I'm not picking on DigiKey here on purpose, just at any distributor, they don't always say constant current. or exactly what I was looking for. So if I could type in on any distributor, like I need a constant current LED driver, that's not something you can really find. But with us, what you would do is you would go in and you'd say, I need a constant current LED driver. And in that case, I needed it to be about an amp. I think we were running at 950 milliamps. So I need it for one amp drive. And our stack was such that we had, I think it was 20, for something volts. So, you know, I need it for a 48 volt power supply so we can buck down to 24. And you could just type that in and then it'll go, it'll set any parameters you need, but it'll also reach into the data sheet and find stuff that is talking about driving a constant current rather than PWM. And so using it for queries like that or... I need a I2C device, but it needs to be not on this address or this address because I already got something there. So that way I don't have to go find all those kinds of parts and then click in the data sheet, try and find it. What address can I put it on? That's the kind of thing we're trying to do. Judy Warner (13:10.215) Hmm. Judy Warner (13:19.921) Very cool. I don't know why, but in my mind, I'm picturing when cowboys do cutting of cows, right? They just sort it out and get the four they want out of the herd of 20. So it really sounds to me like these queries help you just get through a lot of data fast. So it'll give you those options with those parameters much faster. with the technology that is available today, not that's going to be available a year from now. Do I have that right? Collin Stoner (13:55.66) Yeah, and I think importantly, searching through things that aren't parameters on DigiKey or not. So a lot of the stuff that it has, output voltage is this, output current is that, input voltage is whatever. But if I need something like, I need this to be on I square C address, five one instead of five zero. Judy Warner (14:02.001) What do mean by that? Judy Warner (14:17.264) Okay. Collin Stoner (14:18.316) That's not really something that you'll see on the product spec page. But it's definitely something that's in the data sheet. Here's how you set the addresses, and here are the available possibilities. So that. Judy Warner (14:29.384) Mm, got it. So you're pulling out additional information that helps that be more efficient and careful in what you're really asking for. Collin Stoner (14:41.102) Yeah, because you might be looking for a component alternate and you might want to say like, hey, I got this pin out for this MOSFET that I have and it's end of life or out of stock or whatever it is. I think we've all experienced this. But there's a couple different pinouts of MOSFETs. Usually, not always, but, you know, and if you pick the wrong one, it's not gonna work or it's gonna blow up. it's, you know, if I'm looking for an alternate for that and I say, you know, hey, I need the gate to be on this pin and, you know, the drain source to be here. And I need it to be in a package that's compatible with this. That's not something that's gonna be in the specs on the distributor page. That's something you've got to go through the data sheet. And I'm sure we've all done this. It doesn't take forever, but it definitely takes time. And it'd be good to just see, here's a list of options that meet what you asked for. And so it could be. Judy Warner (15:39.887) Right, right. Like I said, it just cuts out your options. It weeds out some options that again, you can use your engineering judgment and intuition by looking through just a handful of things that have already been sorted, right? Instead of looking through everything. And again, you know, when I was with Altium, we used to talk about this all the time as a company that the passion that you all became engineers for is the thing that drives you. And then because we really haven't fully modernized the workflow, like software has done so beautifully, you spend time in bombs and data sheets, like a majority of your time and less time doing the creative fun things that made you want to be engineers when you were kids. So it sounds like you're on that path to make that a reality, which is very cool. Collin Stoner (16:30.924) Yeah, yeah, we want to make the tools that make you faster and spend less time doing spreadsheets because I mean, everyone has to do them, but they're not my favorites. Judy Warner (16:40.679) No. So you're a very young company, about a year and a half old, if I'm not mistaken, but correct me there. And so where are you in your journey? then tell me what kind of interest and growth you've seen as you've gone along this path. Collin Stoner (17:02.67) Yeah, yeah. So we're very young. We were founded in August last year, so just over a year. And for a lot of that time, we've spent nine, 10 months going through and trying to understand how we can apply the AI technologies that are out there and combine them with some traditional stuff so that we can make useful tools and measure how well those are working and then try and find the right fit for the problems you're trying to solve. So we released some initial tools back in June, a couple months ago. And, you know. try to get feedback on how are these working? Do they actually work? And then are they actually useful for you? And we've iterated around that a couple of times now. So we released our search tool for component search back in late August, so right about the one year mark. And we've gotten really enthusiastic feedback from that, more than we were expecting even. So we've told some friends about it and they helped us test. Judy Warner (18:05.127) Hmm. Collin Stoner (18:12.723) And we've talked about it with a few folks at some conferences. And every time we do that, we get this big reaction from... pretty much everyone who's a practicing engineer looking at them like, my God, this is really, this is a really cool idea. I'm surprised someone hasn't done this yet. We actually had a gal take the laptop out of my co-founder's hands. It's like, let me try this. And she starts sniping stuff in there and seeing what it can do. So we've had really enthusiastic growth like that since we released it. And that's been really exciting to see that what I think is useful. is actually useful. Judy Warner (18:50.951) really is. Yeah, that's exciting. basically what you're saying is you've been working really hard for months and now you're sort of beta testing it. Is the tool free at this point or what's your business model as you go forward? Collin Stoner (19:06.158) Yeah, so the tool's free. So if you want to search for parts, you can go to znode.ai and search to your heart's content. And we plan to have that tool be free pretty much forever. So we want to help you find parts. What we're looking to do as a business is monetize a subscription model or some kind of tool that helps you do more advanced things. And the first one we're looking at is finding component alternates, which is, I'm certain the joy of everyone's existence listening. We want to make a lot of tools that make that easier. So finding component alternates is, know, if you exhaust the easy stuff, like let me just change the suffix on here and get it in different packaging, you you got to dig into a number of parts and see are they really truly alternates. And that takes time. Judy Warner (19:37.767) Mm. Yes. Collin Stoner (20:04.133) And sometimes it's not always being done by an engineer, it's being done by someone in purchasing. And then they come to the engineering folks and say, you know, can I use this or can I use that? Judy Warner (20:15.607) All the fun back and forth that you wish you didn't have to do, right? Collin Stoner (20:19.34) Right, yeah, so we're working on tools that let you do more advanced comparisons across lots of parts and lots of the pieces of their data sheets. You can say, like, hey, I've got this chip that is whatever. could be an op amp. And I need it to fit in this package. And I need at least this gain bandwidth. And it needs to be, you know, usable under five volts, whatever else that you need in that circuit. And we're going to surface the tool or the parts that meet that. But then also do things like compare the footprints so you can make sure that their footprints are compatible and show you in the data sheet where those are so that you can kind of overlay them and get a quick sense for this. Judy Warner (21:04.551) Mm. Collin Stoner (21:10.894) things like are the package sizes actually compatible? Are all of the specs that you just listed, do they match? And I air quote specs here because they might not be top line things that you see on distributor. They might be something buried in a table somewhere on page seven. that we have to pull out of there because maybe that's what you care about in this design. That's really the trick of picking alts is looking at the stuff that you care about and making sure that stays the same or better. And then there's a bunch of other stuff that you don't care about, but those things change design to design. And even on the same part, sometimes you put an op amp in that does this. And I don't care that it's got a really big voltage range. I used it because I used it in another part of the design, so I don't care. But some other spots, I might be using it because it's a high voltage op amp. And in those two situations, one for one might not work for the other. And so we need to have a system that can adapt to that. So that's the kind of tool that we're looking to release here possibly early next year. And that would be the tool that we're charged for. Judy Warner (22:20.799) okay, great. Okay, well, Colin, this has been so fascinating and I'm so glad we got a chance to meet and I got a chance to learn about the tool as I'm sure our audience is. So where would you like them to go to learn more about them, about xeno.ai and how would you like them to interact with you? Collin Stoner (22:25.462) you Collin Stoner (22:44.076) Yeah, so we'd love to have your feedback on our tool. We're very young and so is the tool, so we're. We're still trying to find exactly the right fit. And so we'd love you guys to use it. Let us know what you like about it, what you don't like about it. We want to see how you want to use it to make it better for both us and for you. So if you go to www.znode.ai, that's Z-E-N-O-D-E.ai, it'll pop up the search bar. Put anything you're looking for in there. And let us know what happens. there's a little feedback bar on the side or you can can email us. We'd love to hear your thoughts. Judy Warner (23:27.653) Well, again, Colin, so nice to meet you and learn about this tool that you're developing. I think it's very exciting times for the industry and I think your bottom up approach makes a lot of sense. So we don't get way out ahead of our skis by trying to do, to get ahead of where we are right now with AI and LLM. So thank you again for joining us today and I hope you'll come back soon and tell us about the progress after the new year. Collin Stoner (23:54.286) Thank you for having me, I appreciate it. Judy Warner (23:56.487) For our audience, I trust you've enjoyed this interesting conversation with Colin Stone of Zeno.ai. Did I say your name right? Stoner. There's a guy, Kyle Stone, I just did a podcast with. So I'm going to do that over. For our audience, I trust you've enjoyed this conversation with Colin Stoner of Zeno.ai. And I hope you'll go check him out and give him some feedback. I believe in community and ecosystem and Collin Stoner (24:06.243) story. Judy Warner (24:26.245) That's what really wanted me to bring, that's what motivated me to want to bring Colin to you today. So go on over to their website, check it out. I'll put the links in the show notes and we'll see you next time. Until then, remember to always stay connected to the ecosystem. Okay, we've got a little bit of...