Episode Transcript
[00:00:00] Speaker A: Data for the sake of data is not the value. And they need to be much more precise and thoughtful around how they're digitalizing their operations.
[00:00:16] Speaker B: Hello everyone and welcome to the robot Industry podcast. My name is Jim Beretta and I am the host and my pleasure today to welcome Greta Kudalenko to the podcast today. And her company is Asserta analytics and they're here in Canada, they're up in Kitchener Waterloo. And welcome to the podcast, Greta.
[00:00:34] Speaker A: Thank you so much, Jim. Pleasure to be here.
[00:00:36] Speaker B: Can you tell us a little bit about how you got into this industry?
[00:00:40] Speaker A: Yeah, for sure. So I'm a software engineer, graduated from the University of Waterloo at that time, right around when I was graduating. Got very excited about all of the innovation in robotics, but also especially in autonomous vehicles. So when I was finishing up, wanted to really kind of go and explore that area. And so I joined a company called Magna and during that time was focused a lot in the area of, you know, advanced driver assistance systems. It was a lot of fun work that we were doing and really exciting new technology that we were building.
But the interesting thing I've noticed at that time is just how much manual effort it required to test out these systems, show that they were working as expected, monitor them in production, in the field.
And it seemed like a very interesting opportunity for machine learning and AI. And so I partnered with a research group from a University of Waterloo and started looking into that area. And that's kind of how it began. And from that point, you know, we spun things out and started assert us. It's been a great journey.
[00:01:50] Speaker B: Oh, that's wonderful. And congratulations. And it's, it's something that we've needed in the industry for so, so long. So I totally get it.
There are lots of implementers of AI and how are you different?
[00:02:03] Speaker A: Yeah, so a lot of what we're finding in the field is, you know, very generalized platforms. And especially today, AI has often, you know, people started thinking of AI just as large language models. So, you know, everything that anthropic is doing, everything that OpenAI is doing and many other companies, the challenge we see is that a lot of these large language models are really not the best fit when it comes to production because you have very kind of new, a lot of numerical data, a lot of sensory information machinery data. It's not a very stable environment. You really have, you know, product lines changing from around the world, products that they're producing changing hourly, daily. What, you know, is running down one line, maintenance of machines, you name it. There's a Lot of variation there. And so we're finding kind of. And where we focus at Asserta is really in discrete production, where we're seeing a big lack of solutions today just because of the complexity of that type of production and focusing specifically on vertical AI implementation that is fit for manufacturing. So that means it understands the context in production, understands what kind of data can be coming out of those systems. It knows how to tie that through ontology or traceability to the actual results. And what happens to the products at the end, whether they're scrapped, reworked, or even end up in warranties.
[00:03:31] Speaker B: No, that's a. That's a really good answer. You have lots of interesting conversations with your prospects and your customers. What are they concerned about?
[00:03:42] Speaker A: Yeah, so I think a lot of companies are really kind of focusing on two things. One is improving their margins. So continuously finding ways to be more optimal and efficient, especially with the challenges, you know, with global supply chains today and tariffs, companies really need to figure out a way to produce their products profitably and efficiently. And so that's been one thing we hear globally is how do we automate more, how do we make our engineers do more with less, and how do we manage global operations that way? So we see that's one of the areas. And then the second, I think there's a lot of talk in the industry about that AI is going to change everything and everything's going to get faster and there's going to be lots of new revenue streams. And I think a lot of manufacturers are really thinking hard about how should they implement it as part of their organizations in buying those benefits. And so I think that's top of mind for every manufacturing leader today.
[00:04:45] Speaker B: Beyond AI, are there any other big trends that you're kind of noticing or you're seeing with your customers?
[00:04:51] Speaker A: Yeah, I'd say AI is a big one, of course, but digitalization data is still a very kind of fundamental trend. So I think even though everybody's talking about AI, I think the reality in many manufacturing facilities is that, you know, they're still running machines that are maybe extremely old, you know, a decade old.
So figuring out, you know, how to get the data out, how to make it effective, how to not just invest millions into, you know, data collection without that impact on the organization. I think that's been a big kind of shift I've seen in the last couple of years, because even if I look, you know, a couple of years ago, it's been still that primary driver of like, let's collect more data, let's digitalize More of these machines. Let's do all of that. I think it's continuing. But a big trend is now that a lot of leaders are recognizing that data for the sake of data is not the value. And they need to be much more precise and thoughtful around how they're digitalizing their operations. So they're actually enabling AI or analytics use cases for their workforce. So I think that's been kind of one trend I've seen change in the last couple of years.
[00:06:06] Speaker B: Greta, I'm sure you get this question all the time, but what does a project look like? Is there any such thing as a typical project for you? I'm sure there isn't, but I'm just going to ask it anyway.
[00:06:16] Speaker A: Yeah, no, it's a great question because you're right. It really differs by the manufacturer and even within a manufacturer by the site that you go to. So if you take a greenfield site, let's say in electrical vehicles or any EV component, they're very, very kind of advanced lines with newer technologies, newer machines and just very advanced lines. And so you're likely seeing a lot of data come through from those. And so in those type of situation, especially if the manufacturer has an MES system, has a SCADA system that's under 30 day type of implementation, right. You're just connecting to their key storage Systems. Whether it's APIs or MQTT or whatnot. You data is digital. So it's a very quick and easy onboarding with some of our customers that have that, you know, consistent infrastructure, digital infrastructure across their facilities. You know, we're launching in less than 24 hours for a new facility. So if you have, you know, one site in, let's say India and it's using the same MES systems, the same technology as your site in Michigan, 24 hours and it's up but with more Brownside facility, you know, that have diversity of machines where not every machine can be connected. It could really depend. And in those cases, you know, we start with what the manufacturer already has and get that up and running in the first 30 days. Then everything else, new machines, we want to connect new parts of the process that becomes of that continuous improvement. And we try to make sure that if we're asking the manufacturer to invest into it, it's really driven by use case and a problem they have in the facility.
[00:08:03] Speaker B: Greta, who do you deal with at the manufacturing level? Is it the plant manager, is it the owner, or is it the manufacturing engineer or is it everybody?
[00:08:13] Speaker A: Great question.
I think a lot of the times we start to understand the users and the people who are really experiencing that pain. And so that's often the quality engineers, process engineers, manufacturing engineers on the line. They have to make sure they're running production with a certain set of scrap rates, a certain set of first time through rates. They're monitoring the stability of the process. So for them having that visibility is very important. But in every case we have to get that buy in from the plant manager because really, you know, it's an investment and it's an investment towards improving that facility longer term. So it's very important that leadership is bought in. And then outside of the plant's leadership, it's often nowadays, especially in the bigger corporations, you know, we're seeing Industry 4.0 teams or similar digitalization teams or AI teams that could get involved. And it's important sometimes to have them be aware of the projects so that, you know, it can help the global organization. But also at the same time balance that with making sure that the plant is seeing the value quickly so that there's no slowdown in that bringing that value to the plant.
[00:09:24] Speaker B: And you're probably all about the quick wins, right? Like I can't believe how quickly that you can bring a plant online, especially a new plant.
[00:09:32] Speaker A: Yeah, it's always about ROI for a lot of manufacturers, especially these days. You know, the desire to get payback within three to six months is quite high, especially within the automotive industry. And so that's something we're continuously mindful of, that a lot of manufacturers don't want year like years of hypothetical project. They want really AI enabled tooling that can get them to quick wins within production.
[00:10:04] Speaker B: So ROI is one of your first conversations? Likely?
[00:10:08] Speaker A: Always. Yes.
[00:10:09] Speaker B: Well, that's not a surprise. So what is keeping your engineers up at night?
[00:10:15] Speaker A: Yeah, for us it's a, it's continuously figuring out exactly how to get to that value and ROI for our customers.
So, you know, these days, especially in discrete production, as you have, you know, little amount of data from many, many different processes, having things available in real time is very important.
And so that response times and making sure, you know, that the analytics and model predictions are available quickly and point the engineers to the right direction is something we're constantly improving on and working through. And the other thing is really, you know, there's a huge, even though LLMs are not the best choice often for the type of data we're seeing in production, there's still a great way to simplify the interaction between the insights with our users. So we're Continuously looking for ways to incorporate that as an added layer on top to make it easier for the engineers to interact with the data and the insights.
[00:11:19] Speaker B: Greta, at Asserta, you've been very successful with over 400 production lines online.
So that's a big number. And wow, congrats to your, to yourself and to your team.
How do you do it?
[00:11:33] Speaker A: Yeah, it's a, it's a great question and it really kind of. You know, when you, when we chose to work with the automotive industry, the one thing I knew is that it's a global industry. So it's impossible, you know, just to choose to work within Canada. In the Canadian ecosystem you're very quickly are taking globally, across North America, across Europe. It's just that connected. And so from day one we kind of knew that we will need to be able to support our system globally.
So when we were designing the solution, really we were thinking about those two things as scalability and how to make sure that we can easily kind of make the system available remotely across any facility, any production line.
And then two, how to make it easily configurable so that the engineers on the production floor have that usability and can adopt it effectively. And so by combining those two things, we've been able to really scale this while still being a pretty lean team, especially technical team out of Canada. We're still managing all of those deployments globally. We do have customer success, or what Palantir likes to call forward deployed engineers, that type of scenario deployed closer to our customers.
But the solution is extremely scalable and really managed through our Canadian team here.
[00:12:56] Speaker B: That's great. And do you have any use cases that you can talk about? Because I know it's a pretty tricky industry.
[00:13:02] Speaker A: Yeah, no, there's lots. Because we're so focused on making sure that our customers get that impact in production when they're using it. We're constantly looking for ways to augment, kind of add value. And so a couple of use cases that come to mind, one of our customers, they were experiencing a lot of end of line type of issues. So they would produce their product, it would go through multiple stages of assembly, close to 40 different stations per part, and they would get to the end of production and they would fail for noise and vibration. And so that means they have to take the product off the line, disassemble it and basically try again.
And that was impacting quite a big portion of their production. And so what we did is we actually launched our line pulse platform on top of their existing data and then by consolidating data. What we started to be able to do is every bit of information per product. So from across their manufacturing footprint, you know, every process parameter, every machine parameter available, every test result, whether from cameras or a functional test. So basically create that digital fingerprint for every part they were producing. So what that meant is if there was drift or an issue on the line, like a gauge getting stuck or, you know, getting dust on it, we can start detecting that very quickly so they can fix it before, you know, they fail parts down the line. We were also able to have like a closed loop control system where we can recommend to them offsets or subcomponents so that they can optimize the production automatically as it's happening, and that way basically avoid having those downstream failures. So it's really by having this understanding across multiple steps of production, what's impacting the quality of the parts.
That's what our system helped them do. And they reduced their rework rate by more than 65% in a lot of their lines because of it, which was huge for the organization.
[00:15:14] Speaker B: So where does AI fit in all of this?
[00:15:17] Speaker A: Great question. And I think a lot of it is, you know, we're not. The way we look at it was not. We're not replacing the critical thinking that the engineers have to do at the end of the day. But what we're making them is into basically like super people, superheroes in a way. You know, it's that superpower and AI or machine learning based system that can look at so much data continuously and identify those trends or relationships, that is just impossible, like, to analyze so quickly through the engineers. Because if you think about it, a lot of these production sites, you know, they're producing thousands of parts a day, potentially millions a year. It's just so much data. And AI is really great. It's is sifting through all of that information, finding those patterns, identifying those early relationships and calling them out. And so whether that's for cause, analysis, predictive quality, use cases, you name it, we're finding that having that capability to analyze that data very quickly is extremely powerful when coupled with a very strong or even a more junior engineer.
[00:16:26] Speaker B: Greta, can you tell our audience a little bit more about line Pulse?
[00:16:29] Speaker A: Yeah. So LinePulse is a platform that helps manufacturers basically leverage data across their disparate data systems and data sources in production, combine it all in one so that they can understand what's happening on any production line, what's happening to any set of their products, and really help them drive analytics and AI. Capabilities to identify issues early, predict them and really resolve them quickly. And so what we hope manufacturers do is really enable a data driven culture or they can much more quickly reduce their scrap rates, improve their first time through rates and avoid warranties downstream. So, very powerful quality manufacturing analytics system.
[00:17:14] Speaker B: And do you work with automation integrators?
[00:17:16] Speaker A: Yeah, very frequently.
Especially because again, many manufacturers are still going through that digitalization journey. So that means some lines are extremely well digitalized and automated, others maybe partially. And so what we do is of course, when we first integrate, we want to make sure that we're getting the right data into our system and system integrators often help with that. But then the second part of it is once you're starting to analyze, start to leverage your data sets with what you have today, you start to sometimes identify that, hey, here's a gap there, I have that data, I just need to enable that stream and I don't have it. And that's when I think a lot of the times system integrators and automation engineers become much more impactful as well, because it's continuous process of augmenting and enriching the data over time, but in a pointed way so that you're not doing it globally just for the sake of it, but really to enable certain use cases in certain needs of the production team.
[00:18:18] Speaker B: So you now have an AI enabled platform and how does that affect the workflows and I guess data and AI then too?
[00:18:26] Speaker A: Yeah. So one example of course is root cause analysis, something everybody has to do. Every quality engineer, you know, if you have fallout in the plant or warranty issues that your customers are coming back with. So today it's still a very manual process. You do the five whys, you make hypothesis, you do the fishbone diagram, and it's really driven based on, you know, the best knowledge or understanding of your teams. And a lot of the times what we're seeing is that it takes takes days, weeks, sometimes months to go through that process. So an example of how we've leveraged our system is, you know, the data we're getting from production is collected continuously. So let's say you're experiencing an end of line issue or another inline issue or even a warranty case. What you have now is then machine learning and AI enabled models that can look at your failure, look at all of the data you're collecting downstream, and then within just 10 minutes can say, hey, you know, you're station 10 and station 150. They're showing relationships to your failure because something's off in the process there, you know, and it's extremely powerful because by doing that, you're basically not having a team of, you know, five, ten engineers sit in a room brainstorming what it could be. You really have just one person, you know, launch a couple of AI models to really pinpoint where it could be happening, or from a data perspective, where it's starting from. And what that does is instead of, you know, starting with a huge bucket of possibilities, you're really narrowing it down very quickly so that the engineers still need to, of course, go and figure out how to resolve it or how to fix the issue, but you're really reducing the amount of things they need to try before finding that real problem area. So that's been one way that we've seen it incorporated in a traditional engineering workflow in a very effective way.
[00:20:24] Speaker B: So you are kind of, you're creating all this more data driven culture at, for your customers at scale.
[00:20:30] Speaker A: Yes, 100%. Because I feel like that's still where the industry struggles is, you know, there's a lot of solutions out there that are looking to consolidate data, make it available, make it collected. Many manufacturers are storing, you know, years of data, in some cases in their databases. But in reality, the leveraging of that data today is still an extremely manual and challenging part. And we're really looking to optimize that and give the engineers the tooling that they need to make that data actually effective.
[00:21:04] Speaker B: Greta, what happens when something changes in the plant? Like you had an extra station, how does that work?
[00:21:11] Speaker A: Yeah, so that's a great question, especially because most of the time in production, you know, it's impossible that you're not doing anything on the lines. You're continuously, you know, potentially reordering the production steps. You're adding extra sensors, you're, you know, adding extra inline process checks like cameras. There's constantly something changes or even just maintaining the machines. And so what that means is your data is not stationary, it's very dynamic. So the way we handle it is that we developed our system to be very responsive to these dynamic changes. So that, let's say if you renamed your station and moved it somewhere down the line, we can detect that automatically and start ingesting the data automatically from that new location. Or if you similarly added a sensor, removed a sensor, the system will react very quickly to that and make the change without you having to put in a lot of extra effort and manual integration to actually do that, you know, and change your work, your data flow stream. So that's very powerful, especially because, you know, the way we see it is just dealing with the reality of production. You know, you don't have that stationarity that, you know, you have in many other systems.
It's a very kind of. It's a live, breathing organism in a way that's continuously changing itself. And so the way we designed our linefall solution is really to deal with that dynamically as well.
[00:22:44] Speaker B: So, Greta, thank you for that. And who's your. Like, you've got 400 customers or 400 lines, I should say.
Do you have like a bullseye customer?
[00:22:52] Speaker A: Yeah. So we work very closely with discrete manufacturers within automotive. So means, you know, it's tier one producers primarily is who our bullseye customer is, and that's primarily who we're working with today. I'd say from a product mix perspective, we're really seeing a lot of products, everything from, you know, electric motors to axles to, you know, transmissions, engines, you name it. So from that perspective, you know, regardless of what exactly you're producing, if you have a discrete process where you're producing you but at least tens of thousands of parts per year, we see it as a huge benefit to that type of production.
[00:23:37] Speaker B: And how do you engage with a new customer?
[00:23:39] Speaker A: Oftentimes we first try to understand really what is their need and what is the challenge that they're seeing in production. Whether it's their desire to improve scrap rates, reduce warranty rates, or these days even drive up efficiency because, you know, there's a lot of turnover or, you know, workforce is changing. So depending on what that target is, or maybe it's accelerate line, new line, ramp up. We really want to understand what is it that the manufacturer's primary business goals are. And then we start really working away at, you know, how digitalized they are. What can we start with to help them start moving towards accomplishing their goals?
[00:24:26] Speaker B: So how can a manufacturer benefit from linepulse when maybe they have more than one location?
[00:24:32] Speaker A: I think it's extremely powerful, especially if the manufacturer has multiple because, you know, today you might either run them, each location is its own business with little global oversight, or you do have global oversight, but it's extremely taxing. So one thing that we help manufacturers do is drive efficiency in how they're managing those facilities. That means you have a single, single global view, you know, where you can log into any facility, see their data, understand how they're running. And if, let's say you have your one facility in Canada, like say outside of Toronto, that's experiencing a problem that Michigan already knew how to fix. You can basically log in and see what they have done, you know, so it becomes very powerful globally. The other thing we've seen is if you're a manufacturer that's vertically integrated so your one location is feeding parts to another, you can actually enable that global supply chain visibility from a quality perspective. So you can see how did your subcomponents get produced. And if you're having, let's say, you know, in final assembly, some issues, you can then trace them back through your facilities as well to see, you know, maybe it's not coming from your final assembly, but coming from some of your welded parts in another site. It becomes very powerful for that global optimization as well.
[00:25:53] Speaker B: Granted, this has been a great conversation. Have we forgot to talk about anything today?
[00:25:58] Speaker A: I think this has been great. The only thing I would say, you know, AI seems like such a broad set of topics for so many manufacturers today. So I think kind of starting from your target top challenging areas and keeping in mind that some of these things are possible with even limited data, I'd say the big takeaway is really just to get started and look for solutions that have a track record in manufacturing and not just in marketing technology or other use cases. Because really, manufacturing is its own world and it's important to know how to, how to approach it here.
[00:26:42] Speaker B: Greta, when you're not knee deep in projects and data, what do you like to do? Do you have any hobbies?
[00:26:47] Speaker A: Yeah, well, winter's coming to an end, but all through the winter it was snowboarding. I love going up to the mountains and exploring new ones. So this year has been great from a snow point of view. So that's a big one. And outside of that, you know, with the summer coming up, I have a great Weimaraner at.
We're big fans of going to explore different forests. So it's going to be a fun summer for that.
[00:27:10] Speaker B: That's wonderful. And if, let's say there's some engineers or business owners that are out there that are listening to this and they're like, hey, I need to get a hold of Greta or talk to someone from your team.
[00:27:19] Speaker A: How.
[00:27:19] Speaker B: What's the best way to get a hold of you?
[00:27:21] Speaker A: Yeah, that's a great question. Feel free. If, if you guys, if anybody wants to reach out to learn more or see, you know, how we can help your facility, please feel free to reach out to me. It's Gkutulenko AI. You can go on our website to see some great testimonials and case studies as well, or just reach out to me on LinkedIn. I'm happy to have a conversation and see how we can help you think through your issues or even help you solve them.
[00:27:52] Speaker B: Greta, thanks for being a great host and I will put everything in their show notes as well, just in case somebody's driving and they can't to write this stuff down.
[00:28:01] Speaker A: Great. Thank you so much Jim. It was a great conversation.
[00:28:03] Speaker B: Thanks for listening. Today's podcast was produced by Customer Attraction Industrial Marketing and I'd like to thank my team, Chris Gray for the music, Jeffrey Bremner for audio production, and my business partner Janet. And I'd like to send a thank you to Jeff Bremner and you can find out more information on his link tree which is L I N K T R E E G B A U D I O G B Audio Be safe out there.
