In this episode, Eigen Innovations Co-Founder and CEO Scott Everett joins us to discuss the role IoT plays in driving decision-making based on insights, rather than data. Scott speaks to the raw data produced by machine learning and AI technologies and what needs to be done to convert that data into actionable insights truly capable of changing daily workflow. Scott shares the challenges he’s seen working to educate customers on how IoT solutions and AI works and what advice he has for companies who have been struggling with those same challenges.
Scott also shares his experience developing the Eigen Innovations platform and what he’s learned introducing it to customers, as well as the approach he takes in generating meaningful data for customers.
Scott has dedicated his entire career to consulting in engineering and quality control applications. He co-founded Eigen Innovations in 2012 and has been working since that time to bring state-of-the-art technology to the factory floor, specializing in advanced industrial vision and machine learning. Scott is based in Fredericton, New Brunswick, Canada, and spends the majority of his time working with the product development team to evolve Eigen’s AI-enabled solutions as well as pitching the solution to Tier 1 manufacturers around the globe. He’s also in the process of completing his PhD studies in Mechanical Engineering.
Interested in connecting with Scott? Reach out to him on Linkedin!
About Eigen Innovations: Eigen Innovations supports and enhances quality assurance in industrial manufacturing with its unique AI-enabled industrial vision platform. Currently honing in on the automotive sector, Eigen tech has been deployed in several Tier 1 automotive supplier plants across multiple applications (plastic welding, glass soldering, windshield adhesive, etc.).
Key Questions and Topics from this Episode:
(01:02) Intro to Scott
(03:56) What is vision data?
(05:20) Introduction to Eigen Innovations
(07:25) How do you approach conversations about transforming companies from being data-driven to analysis or insight-driven? Do you ever experience pushback against those ideas and how do you handle that?
(11:41) How do you educate companies on IoT and what it can do for them? Do you have advice for other companies that are struggling with that?
(14:16) What were the biggest challenges over the course of developing the platform and introducing it to customers?
(20:28) What’s your approach to generating data that actually changes a customer’s workflow?
– [Narrator] You are listening to the IoT For All Media Network.
– [Ryan] Hello everyone, and welcome to another episode of the IoT For All podcast on the IoT For All Media Network. I’m your host, Ryan Chacon one of the co-creators of IoT For All. Now before we jump into this episode, please don’t forget to subscribe on your favorite podcast platform, or join our newsletter at iotforall.com/newsletter to catch all the newest episodes as soon as they come out.
– [Ryan] Are you tired of overspending on data plans? Do you need more consistent coverage? Are you over negotiating complicated contracts? Well, Simon IoT gets it, that’s why they offer customized, transparent data services across the globe. Flexible contracts, taxes and fees included in one simple price and user-friendly data management, your data is in your control. Their LTE SIM cards are scalable to your needs regardless of the industry you’re in, or the devices you need connected. Learn more at simoniot.com/ifa.
– [Ryan] So without further ado, please enjoy this episode of the IoT For All podcast. Welcome Scott to IoT For All show, thanks for being here this week.
– [Scott] It’s a pleasure, thanks for having us.
– [Ryan] Absolutely, it should be a good conversation looking forward to it. Let’s start off by having you give a quick introduction to our audience. Talk a little bit more about your background experience and kind of what led to the founding of your company.
– [Scott] Yeah, absolutely, so the story of Eigen, I’m a mechanical engineer by trade and we’re actually a spin-out company from the University of New Brunswick here in Eastern Canada. And really what we’re focused on is advanced control and quality control solutions for industrial manufacturing. We do a lot of work with vision systems, so we’re an industrial vision platform. And really the company was born out of a really simple observation. I was working on my masters in mechanical engineering and we were engaging with a lot of different customers. This was back in 2010, before all of the hype around AI and IoT really took off. So what we saw time and time again, when we went into factories is there’s an immense amount of knowledge that operators build up over time and it’s intuitive. And so a lot of the way that they’re managing their factories and controlling their machines was based on this intuition. And we really just saw an opportunity with the data that was on the machines and the capability to capture new data, to say there’s gotta be a better approach that makes it a lot more optimal and really takes the variability and the ad hoc nature out of it. So it was a really interesting time because as IoT started all of the technology around basically capturing data and the new exciting information that could be captured and networked and aggregated together, we just saw a ton of opportunity. So that was the concept for the company. We really got into vision data because it’s such a rich and interesting source of information that can augment a lot of what’s already being captured in factories. And so we started early on working with the technology of machine learning and artificial intelligence. And back in those days nobody had any idea what AI was, right? So it was a really interesting journey to introduce that to a customer base and then figure out how to help them, understand the technology and understand what its impacts could be. So fast forward eight, nine years later, and AI it’s proving that it has immense capability, but we’re still, I think in general IoT there’s still a long ways to go to recognize its full potential, right?
– [Ryan] For sure. So when you mentioned vision data before, what is that exactly? Can you explain to our audience kind of what that does?
– [Scott] Yeah, so is basically cameras that are installed on factory lines. We actually work with different types of cameras. So you’ve got regular cameras that are taking pictures of charts. And you’re able to with artificial intelligence actually automatically detect defects that has been a manual inspection for a long time. And so manual inspection, there’s a lot of variability, a lot of subjectivity when you have a lot of different people trying to determine whether it’s a good or bad quality product. So the camera technology allows us to capture that richness of information and start to create standard strategies for quality detection. What we also found that was really interesting is we work a lot with thermal cameras. And so it gives a whole different spectrum of data to capture about the manufacturing process. And what we found is that type of data can be really, really beneficial in helping see how things are changing so long before you actually make a bad part. You’re seeing these trends and you can start to correct for that, and I think that’s where the real optimization and efficiency can come from.
– [Ryan] Absolutely, can you share any potential use cases or applications of your technology kind of in the real world that you’d be comfortable kind of giving some insights into our audience of how it worked, maybe the story behind kind of what was the problem initially? What did you all, and then what role did you play in helping solve it? Just so we could kind of put a real world situation to it all.
– [Scott] Yeah, absolutely, so for us we’re very focused on the automotive industry. They have a lot of high volume, high value parts, right? And like one use case that’s a really interesting one are the headlights and taillights on vehicles. So if you go back and you look at like a Toyota Tercel, from the 90s or whatnot, everything was very square, very simple headlights back in those days today, you look at the design of a vehicle and a headlight is a major styling element on a car and it’s a safety component and there’s a lot of new embedded technology that’s going into these components. So the value is quite substantial. And one of the big things with the safety component, like a headlight is it has to be perfectly sealed, right? Or else it’ll fog up and cause a warranty recall or whatnot. So what we actually do is we’re actually capturing images of every single part and connecting that to the data that’s coming from the machines. And we’re actually able to verify that every part is gonna be properly sealed. And like quality testing today, that’s usually done offline and it’s an ad hoc kind of process, so we’re actually able to give a hundred percent traceability and guaranteed quality for every single part. And so safety critical components on vehicles, a lot of plastic welding, a lot of injection molding, those are the types of use cases where we’re really combining the richness of the vision data that we capture with all of the process data.
– [Ryan] So when you speak with organizations, there’s been kind of comments made about needing to transform an organization from being data-driven to kind of insights and analysis driven. How do you kind of approach those conversations with organizations that you speak with, which oftentimes when you get into manufacturing and that kind of industry, more of the industrial side, you can be met with a decent amount of pushback and kind of hesitation to adopt new technologies like this. So how did you all kind of approach those conversations? And how do you just add it at a high level feel that IoT applications can help people and organizations achieve that transformation from being more data-driven to that insights and analysis driven?
– [Scott] Yeah, it’s really interesting. I mean, in our journey, what we set out to accomplish, eight, nine years ago was really this optimization of the manufacturing process and guaranteeing that you’re making good parts every single time, increasing efficiency and that is the holy grail for all manufacturers, right? They wanna make more product, waste less, be more efficient. And what that really boils down to, is helping manufacturers have a specific answer to, okay, there’s been changes in the conditions in my factory and it’s starting to create bad parts, what do I do? That’s the big question that they need to answer in the moment and in real time. And engineers are always struggling to kind of adapt and compensate for the things that just naturally change. And what we found is you really have to get to that answer to provide value to the company. So what we’ve observed is in the early days of IoT, it was really all about capturing data, right? And so it was all about the devices and networking and getting data to a centralized spot so that you could actually start to use the data. And a lot of those use cases are always focused on the data and there’s sort of this afterthought of, oh, and once you get all the data then these insights will just magically appear. But that’s not the case, right? Data starts to become very overwhelming. And so I think some of the pushback that happens in the industry is the fact that once all this data starts flowing in, it actually creates more problems than it solves, and it takes a lot of time and energy. So we’ve seen the opportunity now that there’s an infrastructure like IoT to really create the conditions. In today’s world is really about focusing on how do you extract those insights and how do you take that information and have it changed your day-to-day work, right? So if you’re not actually making decisions and adjusting your practices in a real-time basis, based on the data, then the data is just a distraction. So you’ve got to get to that. And I think one of the critical things is really understanding their world, being very empathetic to what their day-to-day looks like. And so if you can’t speed up and give them valuable insights in the moment that they need to solve the problem, the pushback comes from the investment of time and energy, to do all this work in building an IoT solution. And at the end of the day they’re still falling back on the way that it’s always been done, right? So one of the things that I find very interesting is if you start to dissect, if you go from the other side and say, well, what are the things that they can change? What are the decisions that they can make in the moment and then start to plumb back, well, what information would they need to have to feel confident to make a change or to do something differently? And it really comes down to building a story out of the data that’s very, very easy to interpret and really get your mind around, this is what the data means, this is what I need to do.
– [Ryan] That’s fantastic, thanks for kind of elaborating on that. And I think one of the things we noticed a challenge way back when we started IoT For All was just on the education of what IoT is, what AI is, kind of how things work in the space, all the way down to the fundamental, like technologies that are involved. In your side how big of a challenge is that on the education of what IoT and how AI works? What things can do for their business, all that kind of area, like what do you all do? How do you approach that? And just generally, what advice do you have for other organizations out there who may be struggling with that?
– [Scott] Yeah, I mean, when we first showed up to factories and started talking about machine learning, it was a lot of blank stares. So we had to learn how to actually give a primer on what machine learning actually is. And you gotta remember you’re coming into a world where automation is very pervasive and the concept of automation is there’s rules and very specific logic and you set those up and ideally you can set it and forget it, right? And the connection of the rule to what it changes in the processes is pretty easy to understand. What machine learning is really powerful is in those areas where that logic and those rules, that it falls apart, right? Where the complexity of the process, it’s difficult to create a rules-based strategy and maintain it. And so AI can learn these really complex patterns, but you gotta realize that it’s in that learning process it takes time and it takes some mentorship, right? The beauty of it, is it really helps you to understand and find the variability that you might not be paying attention to, but it does really require a certain understanding that this is an iterative and evolutionary process where as things change these algorithms are gonna learn new things, and so there’s an engagement that has to happen from the team that’s managing the system to keep it dialed in and making it effective, and actually what we found is that’s the process that we help manage for our customers, is our platform is really about how do we make sure that we’re contextualizing the data and keeping those algorithms dialed in at all times.
– [Ryan] Absolutely, and when you went through the process of kind of developing the platform and your solution, and kind of offering to the market and working to make all this data kind of understandable, giving the customers really greater control over how things are working, how things are performing, what challenges did you kind of encounter with that process? And I guess what big learnings were you able to kind of take away from it, to kind of get where you are now?
– [Scott] In our world, which I think is common in a lot of IoT scenarios, the data that you’re capturing is usually very specific to the application, and the diversity of the data, you can capture a lot of information, but the diversity in that data can be fairly minimal. So the applications where AI is really great is where you have some sort of standardized data input at volume, and you can train these networks. So when you start to break down to the specific use cases, particularly within manufacturing, it requires a different strategy. And one of the challenges I think for us that we’ve had to develop solutions for is the contextualization of data. So something, a pattern that you recognize in one circumstance might not mean the same thing in another circumstance, right? So if you change a material, for instance, the data that you capture for one type of plastic might not directly apply to another type of plastic. And so then what you really need to take care is how are you grouping and performing your analysis on that data. So that the context, like naturally as humans we’re always adjusting to our context, right? And so it’s the metadata, or all of the information around the core data that’s captured that really helps contextualize it and that’s really, really important for creating consistent insights off of this information. I also think that one of the things that’s a real challenge, there’s a lot of talk about standard IoT architectures, right? And it’s pretty well understood now what an IoT architecture, there’s a lot of templates and patterns. And so it’s easier to describe the components of an IoT solution to customers. The more difficult thing is there’s no standard architecture for insights, like what is an insight, right? The way that I think about an insight is, an insight is a story that gets you to a place of confidence in making a decision. And the challenge when you have fragmented data and you have data that’s very unique to all these applications is how do you create a consistent architecture of generating insights that lead people to that decision point? And so I think it’s a really interesting time, we’ve got all these data scientists now that are going into the data lakes and trying to create, value off of the data. Most organizations don’t have data scientists that are able to really provide that in real time. So in terms of scalability of value for IoT solutions, you’re getting to a place where there’s a standard way that you can actually step through and create a very explainable sequence of logic off the data that gets them to the decision they need to make. That’s where I think the scale potential is still, still it’s a blue ocean because we’re now with the maturity of the data infrastructure, we’ve got all this fantastic information. We’ve got to create the scalability of the insights within an organization, right?
– [Ryan] Yeah, I think one of the biggest challenges I’ve seen a lot of companies have is the ability to build a solution that’s very technical, but allow it to be used by non-technical individuals who can understand the information. They can go into an industry without having to require the company to hire new people, to understand how to use the system, but can apply it to their every day process. So that the end, whoever the end user is they can build for that end user and not over-complicate the role or kind of cause any problems in kind of what they already have going on. And I think that’s a very unique attribute for a company to be able to do and do well.
– [Scott] Oh, I a hundred percent agree. When you think about the world of the people who are boots on the ground and are the users of IoT solutions, oftentimes data analysis and everything that we’re talking about is a completely new job, right? We’re introducing new work for them that requires that training. Really, for it to be a value I think we have to augment and trigger off of the jobs that they’re already doing and find ways to make those jobs way more efficient and way faster. And so thinking about it from that perspective of saying, okay. For me explainability is the inverse of data science. And so a data science approach is I’ve got all this data, I’m gonna start wading through it, I’m gonna apply different techniques and hopefully at the end of it I come out with something that gives me insights to go do something different. But an explainability paradigm is you audit, you use the artificial intelligence to generate the answer, but then you still can take the responsible humans and take them back through that data and tell the story. So you’re saying, “Hey, I’ve got something that’s gonna help you do your job easier and faster, here’s what I’m recommending.” And for you to feel confident in making that decision, here’s the logic and here’s that journey back through the data, but it’s often the reverse in terms of the way data science has been approaching the problem.
– [Ryan] Yeah, that makes a lot of sense. I think that’s a really good way to kind of put it and think about how things have been done in the past and how things need to be done in order to be successful. And kind of to elaborate on that before we wrap up here, I did have one other question I wanted to ask you, when a lot of these companies are being pitched on IoT solutions, and they’re kind of sharing why their IoT solution is a perfect fit, a lot of it is focused on the new data that can generate insights, but rarely are those insights being generated in a way that really changes the customer’s daily workflow. And I feel like that’s kind of explains what you all are trying to solve and to do. And if can you elaborate a little bit more on kind of how that has sort of been done and approached and your kind of take on it, which is kind of the opposite, which is kind of showing how it can really change the customer’s daily workflow as opposed to just focusing solely on and look at all this data we can generate for you.
– [Scott] Yeah, absolutely. So in our world of manufacturing IoT applications, a lot of testing on the product itself it’s destructive, it’s offline, it’s very, very labor intensive. And that’s where a lot of waste and costs, like quality control is really, it’s a cost center, right? Like if you didn’t have to invest in it, manufacturers just wouldn’t. And one of the things that has been really exciting I guess for us, is what we unlock in terms of the data, instead of just testing like 1% of your product and destructing that and throwing it out as waste, if we can capture the data in process, inline, quality control is a lagging indicator, you detect the problem after it already happens. If we can actually create leading indicators of quality, we can actually drive to zero defect manufacturing where you don’t actually have to have the same operating procedures around all of this destructive testing. Destructive testing is usually your random sampling, and you’re hoping that you capture something that’s representative of the process, but when you have the inline leading indicators, you can then be very targeted in what you actually test. And it’s usually when there’s some new variability, all of the rest of the product, you can actually certify that, you can help them guarantee that, that product is a good product and you don’t need to destruct it. So when you can change the operating procedure of an organization that’s where the value really starts to stack up, right? It not just, “Hey, let’s go in and try and find some data.” It’s like, okay, fundamentally, what are the processes that can be changed that are significant value for your organization? And then work back from that to say, “Okay, our goal now is to actually change that procedure,” And it transforms the organization. So it took us time to get to that point but I think that’s the really exciting thing is when you can start to enable people to do things differently, faster, better, more efficient, that’s what we need to get to. Okay, we’re gonna show you some fancy charts, and you figure it out from there, right?
– [Ryan] Yeah, I think that’s all part is often very much overlooked. And once we start building from the end user backwards, I think that’s where you start to see success. And you also have a very good grasp on it. So if anybody out there is listening that’s kind of wants to learn a little bit more about what you have going on, kind of follow up from this discussion, any questions, that kind of thing, what’s the best way they can do that?
– [Scott] If you wanna find out more about Eigen Innovations you can find this at eigen.io or on LinkedIn, and we’d be happy to chat about manufacturing applications. We are focused a lot on parts manufacturing, plastics, all different types of processes. So love to chat, really talk about what your requirements are for an insight architecture, and then work back to the IoT solution, yeah.
– [Ryan] Fantastic, well, Scott this has been a real pleasure and a great conversation. Thank you so much for your time and being here today, we’d love to have you back at some point in the future, talk more about what’s going on.
– [Scott] Sure, we’ll really appreciate it. Thanks for having me on and yeah, we’ll chat soon.
– [Ryan] Awesome. All right, everyone, thanks again for joining us this week on the IoT For All podcast, I hope you enjoyed this episode. And if you did, please leave us a rating or review and be sure to subscribe to our podcast on whichever platform you’re listening to us on. Also, if you have a guest you’d like to see on the show, please drop us a note at [email protected] and we’ll do everything we can to get them as a future guest. Other than that, thanks again for listening, and we’ll see you next time.
PlatoAi. Web3 Reimagined. Data Intelligence Amplified.
Click here to access.