AIP Podcast EP 78 - AI-Powered Manufacturing Quality with Acerta’s Real-Time Root Cause Detection

This episode’s guest, Greta Cutulenco, Co-Founder and CEO of Acerta, joins host Anne to explore how AI is transforming quality and decision-making in manufacturing. Greta shares her journey from automotive engineer to AI founder, and how firsthand experience with slow, manual quality processes inspired her to build Acerta’s LinePulse platform. She explains how combining machine learning with production, process, and quality data creates a “digital fingerprint” for every part, enabling manufacturers to detect issues in real time, reduce rework, and prevent costly defects. Greta also dives into real-world results, including major improvements in scrap, throughput, and root cause analysis for global manufacturers. Plus, she offers a practical perspective on the future of AI in manufacturing—why human expertise remains essential, and how AI is augmenting, not replacing, the workforce. Tune in to learn how Acerta is helping manufacturers unlock faster insights, better quality, and smarter operations at scale.

Find out more here: https://acerta.ai/case-study/solving-multi-plant-quality-issues-at-dana/

Follow AIP Affiliate, Acerta
Website: https://acerta.ai/
LinkedIn: https://www.linkedin.com/company/acerta-systems-analytics/

Follow AI Partnerships
Website: https://www.aipartnershipscorp.com/
LinkedIn: https://www.linkedin.com/company/aipartnershipscorp/
X: https://twitter.com/AIPartnerships

The AIP Podcast is hosted by Anne Cheng, on behalf of the AI Partnerships, a Railtown company

Transcript

SPEAKER_00 0:00

The world of manufacturing sees the generation of copious amounts of data, much of which is manually passed and analyzed to reveal the true source of waste leakages and deficiencies, leading to operators being unable to identify and respond to issues as they happen. This caused manufacturers losses through recalls, rework, wasted time, wasted resources. And our guest today is attempting to leverage AI to get data-led decisions to operators quicker. Saving countless heart takes. Hello, good morning, good afternoon, and good evening to our audience from wherever you are tuning in from. I'm mentioning the whole self-year AIP podcast, and thank you for tuning in to our episode today. I know we've done this before, the world of manufacturing, but remember it's a vast structural beam in our daily lives, from the food you eat to how you communicate to how you get around. But every manufacturer will tell you how difficult it is to solve for line-related problems. Call it Six Sigma, call it root cause analysis, and every other framework that's been emergent, you simply can't shut down a line when you're in production because every second counts. But this leads to digging yourself deep into a hole a lot of the time, especially when issues arise and they get prolonged or even perpetuated. Today's guest, Greta Kunalenko, is a founder of Acerta AI who is working to solve line-related issues. Greta, thank you for being on our show today.

SPEAKER_01 1:53

Thank you so much. Excited to be here.

SPEAKER_00 1:56

Tell me about your origin story. How did you come to become one of the founders of Acerta AI?

SPEAKER_01 2:03

Yeah, so um I was a graduate of uh University of Waterloo way back when. So started my career as an engineer. Uh, when I was uh finishing, I was really excited about automation and robotics, um, and especially in the automotive industry. Uh, right around that time, it was just so much talk about autonomous vehicles, autonomy, how cars are basically like computers on wheels today. So that got me very excited, and I decided actually to pursue a career um as an engineer in automotive. So I joined uh Magna, one of the largest um tier one manufacturers uh in Canada, um, and especially their electronics division. So looking at a variety of different subsystems that are enabling ADAS capability today. So the driver assistance systems like backwards facing cameras, um forward collision alerts, things like that. So I was working there and uh it was great. Um, but one of the things that I found fascinating during my time there was that manufacturers, even as big as Magnet, they're running, you know, almost 300 facilities globally, huge company. Um, they were still using, uh, we were still using a lot of manual approaches when it came to quality. So if something wasn't looking right, uh, if we needed to figure out how something could fail in the field, I had to think through, you know, our FMEAs, so like the failure modes analysis, look through all of the ways and think through how it could fail, how it could impact someone. And it seemed to me like a very slow and laborious process. Um, and I was doing kind of um part-time masters at the time with the researchers at the University of Waterloo, and we were looking at especially at AI applications for you know, industrial systems, and you know, cars were one of those potential systems since they're so safety critical. Um, and that's where something clicked, you know, um, just given how much data is available within the industry, why not leverage machine learning and AI more to change how quickly engineers like myself could identify quality issues, deal with quality issues, and really help a major company like that uh prevent and um completely avoid this type of problems inside of their own production, but as well as in the field. And so in 2017. That's amazing.

SPEAKER_00 4:41

Sorry, go ahead.

SPEAKER_01 4:42

Yeah, and so in uh 2017, um, myself and two of the researchers from the university, we decided to spin off some of the research that was being done and uh start what is now Acerda.

SPEAKER_00 4:56

That's wow, um that's really uh impactful. But let's talk a little bit about Aserta. I got so excited hearing to uh your backstory. Let's tell me what does Acerta do and how does it do it? How does it solve the problems, especially along the manufacturing process? Especially, you know, we know that production, when it's taking place, it's virtually impossible to stop.

SPEAKER_01 5:21

Yeah. Yeah, no, that's a great question. So um Acerta really kind of came from that original vision that there's so much data available today, but the challenge is that manufacturers often just lack the tools to really be able to harness and leverage that data for day-to-day decision making. Um, and that's honestly still true today. A lot of them use um minitab, you know, uh Excel, just visualization tools, which are great when you have a little bit of data here and there. But when you're really thinking about all of the workflows that quality, process, manufacturing engineers are doing in production, um, that data really should be there as part of the day-to-day decision making. And so um at Acerta, we've created a platform called Line Pulse that's um really the unique thing that we do is we combine data from quality as well as process and machinery together to then help create uh basically like a digital fingerprint for every single part you're making in production. So whether you know it went through torquing, pressure send uh pressure stations, you know, um die casting, final testing stations, visual inspection. We get all of that little pieces of data to understand what's happening to the product as it's going through production. And that way, then applying AI and machine learning algorithms to help identify issues as early as possible, alert them to the engineers on the line, get them early visibility if their process is deviating from expected performance, um, and performing really predictive uh analytics to then highlight, you know, if some variation can then drive increase in scrap, increase in rework, or even potentially uh warranties in the field. Um, so from that perspective, we're really changing how the quality workflow is done by basically enabling the engineers with data and with the tools they can use to really accelerate their decision making. So that's been um a big focus of what we're doing with Asserta. And again, we're seeing a lot of success with that in the industry, especially as a lot of digitation is happening in production and more and more data is becoming available. There's just no other way to run your factory today, in my personal opinion, in the 21st century, without making data part of your day-to-day workflow.

SPEAKER_00 7:56

That's amazing. But let's start to make it real for our audience. Perhaps you could share a use case or uh one of the customers that you've worked with and how it's become what what you've successfully deployed, what did it mean and translate into for that particular customer?

SPEAKER_01 8:15

For sure. Uh so one of our early customers that we were working with is uh Dana International. So they're a major manufacturer of axles, drive lines, and similar systems. So basically a supplier to the automotive industry and broader commercial vehicle industry as well. Um, one of the challenges we started looking at with them, so you know, as a very progressive manufacturer, they've deployed a manufacturing execution system in their facilities. They were deploying SCADA systems to manage and control production. So it was a very good mix of both automated but also manual process steps that they were basically collecting data from. And just to give you a sense of what a typical assembly line is, you know, you're doing step after step after step where you're taking like a couple of pieces of metal, adding them with each other, adding more things to them, basically assembling into where each step you're adding a little bit more value. And at the very end of production, of course, they're doing testing. So they're checking, you know, is it producing too much noise? Is it the right amount of noise? Is everything moving the right way? And so if you can imagine lots of data, like we were looking at uh, you know, a couple of hundred parameters tracked within production, number of parameters tracked about what type of product they're producing, because you know, they're making lots of variants continuously on the same lines, and then a huge amount of testing data. So the challenge they were facing is that as all of this was happening, everything was in spec, things look good, you come to the end of the production and you're failing your parts because of too much noise or because of something else. So they have to take them off, rework and try again. Big impact to the manufacturer, you know, all of that uh waste as well as just loss of uh time and and labor effort. So, what we did, uh we deployed line bulbs all across their production. And so we were constantly in real time monitoring the data as every single part comes through. And you can imagine, you know, they're producing uh a part every couple like every minute, every couple of minutes. So really big volumes continuously, you know, moving the line. And as you said, cannot stop it, have to keep going. So if you're producing something bad, you just suck it up, take it off, rework it, try again. So um, what we started doing initially is looking at that data to see, hey, is the process performing well? So is there variation? Are there changes or gauges being stuck in production? Are there torques that aren't quite reaching the right amounts? So we started giving them early indications if there was process changes that may not have, you know, crossed the line for their um expected controls just yet, but are varying enough that it's impacting their final production. And we found we started finding for them very early indications, for example, if certain measurement equipment was getting stuck, so it was reporting incorrect values that they were later used in production. And so that way they could fix it much earlier. The other thing we did is through machine learning, we actually optimized the process. So they have some subassemblies that you know require really precise pieces of um pieces to be put together for every for the whole system to work. So we started looking at how we can optimize that automatically and provide the right offsets, the right subcomponents so that that whole process works right the first time around. Um, and so through all of those capabilities, then also automated root cost capabilities where they can look backwards and see what in the process is causing some of their problems, all of that has started making major impacts across their organization. So they've seen um 65% improvement in rework rates, significant double-digit improvement in scrap rates, um, throughput rates going up by 8% in some of the locations. So really uh good results by leveraging this across the organization. Um, and last year we even took it a one step further, actually. So they their they have connections between facilities where some of their, let's say, um Mexico or North American facilities are supplying European facilities or vice versa. So there's this like connected supply chain within their organization. And we're now able to even monitor their their whole supply chain. So if you're producing a part in Fort Twain and shipping it to Toledo, uh, we can start monitoring that path and what happened with that part that ultimately reached Toledo. And so if any issue originates that's coming from a supplier plant, we can now drive analytics across the organization so that they're becoming better not just within any one facility, but across their whole manufacturing network. And that's been really powerful.

SPEAKER_00 13:31

That's amazing. Um, it looks like we're almost out of time, but I I really wanted to, you know, do a small final question. Um, especially because, you know, today we've got artificial intelligence really solving a lot of the problems that we see in many manufacturing, um, even automating almost every single process. Do you think that means that we might not even have a manufacturing industry or humans in the manufacturing industry tomorrow?

SPEAKER_01 14:07

That's a great question. Um, find that the automotive industry is often at the forefront of innovation in manufacturing just because it's such a high pressure, so much, you know, competition, low margins. So there's a lot of pressure to automate and get better. Um, I do think that a lot of automation is coming to that industry. It's already been happening, you know, a lot of automated camera systems, robotic systems. So a lot of uh work is being um done more efficiently. I do think there's still a lot of very intricate processes within manufacturing where collaboration between robotic systems and AI systems and people is really important, however, because we're seeing, you know, still there's you know, take a foam manufacturer, for example, those folds and things inside of your seat, it's so hard, even till this day, to do that visually that you just have to use people and that touch and that precision um to really deal with it. So I don't fully see it getting completely automated. But the other thing I also have seen is that um whenever there's an issue, you know, you can optimize your robot, but in assembly facility or in majority of manufacturing, it's not just one part, one process, and you're done. It's like step by step by step, multiple things that have to be done. And so even if you automate one robot perfectly, once you get started on the whole system, you know, when things go wrong, you need engineers who understand what you're designing, what the process is doing, um, how to think about it really to get involved. So I do think it's a it's a collaboration long term. Um, you know, there's of course that North Star of lights out facilities everywhere, but uh what we're seeing even in the most advanced facilities today, there's still, you know, those uh critical engineers that are monitoring and observing and resolving problems when things don't go well. So I think that that's where it's gonna continue moving towards over the next decade or two.

SPEAKER_00 16:14

Thank you. Greta, this has been superbly insightful and thank you for such a wonderful session. I really enjoyed myself and I hope you have to. To our audience, once again, thank you for tuning in. Please don't forget to like, share, and follow us. And share this with somebody who would benefit from it so that we can continue to bring you more AI content that's really changing the world. Once again, my name is Ancheng. And on behalf of the AIP podcast, my guest has been Greta Kodalenko of Azurda AI. Thank you for tuning in.