Vision Podcast #27 – Designing for Longevity w/ Armando Bompane, ATEK Engineering

Bridging the Gap Between Manufacturing Challenges and Machine Vision Solutions

MVPro Media – The Vision Podcast #27

Guest – Armando Bompane, President | ATEK Integrated Engineering

Armando Bompane, President of ATEK Integrated Engineering, joins Josh Eastburn to discuss why designing automation systems for longevity creates more value than simply delivering projects on time and on budget, and how long-term thinking shapes better engineering decisions.

In this episode, Armando explores what it means to develop automation systems with a decades-long mindset, from designing for maintainability and future expansion to balancing performance, reliability, and lifecycle costs. The conversation covers engineering best practices, modernization strategies, customer collaboration, equipment upgrades, and how manufacturers can make smarter investments by planning beyond the initial installation. Armando also shares practical insights on building automation solutions that continue delivering value long after commissioning.

---

On this page:

• Podcast player
• Guest information
• Useful links
• Episode chapters
• Episode transcript

---

Listen to the Episode:

---

About our Guest:

---

Useful Links:

Armando Bompane: linkedin.com/in/armandobompane/

ATEK Integrated Engineering: atekintengineering.com.au/

ATEK Integrated Engineering on Linkedin: https://www.linkedin.com/company/atekintengineering/

---

Episode Chapters

Click onto the chapters to access the relevant sections of the transcript below.

1. From Manufacturing to Machine Vision
Start: “I love it how you started that question, sign up for the pain…”

Armando shares how his career across Australian manufacturing, European machinery, and North American automation led him to found ATEK, with the goal of helping manufacturers bridge the gap between production challenges and machine vision technology.

---

2. What Makes a Good Machine Vision Project?
Start: “There was a project that I got involved in where the OEM had sold some robotics that was integrated with machine vision…”

The conversation explores why successful vision systems begin with defining what quality actually means, and how translating subjective inspection criteria into measurable, machine-readable data is often the hardest part of any project.

---

3. Beyond the Camera: Understanding the Machine Vision Stack
Start: “I think the biggest gap in understanding by end users is that they think machine vision is really about a camera…”

Armando explains why machine vision is far more than a camera, breaking down the hardware, software, optics, and engineering decisions required to build reliable inspection systems that perform consistently on the factory floor.

---

4. The Gap Between Smart Cameras and Custom Systems
Start: “I have this conversation a lot, and I’d love for people to give me feedback on my mental model of the market…”

Armando shares his view of the machine vision market, explaining where smart cameras deliver value, why many systems become underutilized over time, and how the next generation of vision platforms must balance simplicity with long-term flexibility.

---

5. AI, Explainability and Engineering Trade-offs
Start: “Yeah, I think the AI conversation is probably a podcast in itself…”

Josh and Armando discuss where AI adds the greatest value, why traditional rules-based vision still has an important role to play, and how combining both approaches can produce more reliable, explainable, and practical machine vision systems.

“The end user sees it as a camera, but machine vision specialists understand that there’s this whole hardware-software stack involved.”

---

Episode Transcript:

Armando Bompane – ATEK

Fundamentally, why machine vision is a challenging field is because the end user sees it as a camera, but machine vision specialists understand that there’s this whole hardware-software stack involved.

Josh Eastburn – host

Welcome to the MVPRO Podcast. There’s a question that comes up early in many machine vision projects, and it sounds deceptively simple. What is a defect? Not in a philosophical sense, but in a precise, measurable, machine-understandable sense. My guest today is Armando Bompane, the founder and engineering solutions director of ATEK Integrated Engineering, an Australian engineering and technology company specializing in machine vision systems, industrial AI, intelligent inspection platforms, and automation technologies for manufacturing industries. He founded ATEK to focus on custom machine vision and automation solutions where the environment is challenging, the products are variable, and the stakes are real. ATEK’s work covers food processing, packaging, agriculture, and industrial manufacturing, with a particular emphasis on delivering practical, deployable vision systems that hold up on the factory floor. Without further ado, please enjoy my conversation with Armando Bompane.

Josh Eastburn – host

I think with a system integrator in particular, right? Like that fills a particular niche in the industry that I think is really interesting in between the end user and the supplier and all of that. And I always wonder how does somebody end up doing that as a job, signing up for that pain, however you want to think about it. How did you end up, let’s say, let’s just talk about ATEK. How did you end up starting ATEK?

Armando Bompane – ATEK

I love it how you started that question sign up for the pain. That was a really, really good way to ask the question. Now look, my background, essentially I’m a mechatronics engineer, computer science grad, and I’ve always been fascinated by science and engineering ever since I was a little boy. And it turns out that back in the early 2000s in Australia, when I was studying engineering, we had a very, very large automotive manufacturing sector here. And then there was some changes in that sector. It started to decline somewhat. And I, I moved into the packaging sector. The packaging industry was booming here. We have a large food manufacturing base, a large agriculture and food manufacturing base. And I sort of moved straight into manufacturing of packaging products. And I was actually working in the plant for a number of years, you know, learning production systems, learning manufacturing systems, lean and things like that. And, uh, and then after that I, I moved into the, uh, the OEM side. So I learned about manufacturing, and then I started talking to manufacturers about what kind of equipment could make them more efficient, could reduce their cost of goods, could, uh, you know, help them package products, process products, and things like that.

Armando Bompane – ATEK

So I was actually working for a, uh, a German company, uh, as an— which was an OEM packaging machine company, and I learned a lot about European way of of building machines, uh, not only German way of building machines but the European style of machinery and integration and automation. And then after that, I actually moved over and worked for an American-based company doing the same thing. So I was learning a lot about how North American factories were building equipment and automation. And with that, I learned a lot about sort of— I summarized my career in like 3 points. There’s Australian manufacturing, then there was European machinery, North American machinery, And then I took the lessons from all that and I decided to, to found ATEK essentially 3 years ago. And there was a big, a big need in the market, the Australian market, to fill a gap around helping manufacturers here become more, more efficient, more effective, particularly food manufacturing and agriculture. There was a gap between the OEMs from the other side of the world and, and what was actually being communicated and supported here in Australia. And because I had this global experience, I thought I’d bring that to the table here in our home base.And that’s how I founded ATEK. And that’s essentially a bit about the journey and how I got here so far.

Josh Eastburn – host

Yeah. What regions are you covering now with Australia as your home base?

Armando Bompane – ATEK

We’re based in the north part of New South Wales near the Gold Coast. And we service all of the eastern side of Australia, which is pretty much where most of the population lives. So Melbourne, Sydney, and Brisbane. But we have a very strong regional footprint as well. So in Australia, regional manufacturing is still very strong, country towns employing people from the local region. So we, we make an effort in actually supporting the, the factories in the regions and putting a large focus on those factories because sometimes they lack that ability to get that kind of support. So that’s where a lot of our focus goes, but we’re also building systems and exporting them to Europe and, and North America as well. So we are quoting customers in Europe and, and North America, and we are trying to grow into those sectors as well.

Josh Eastburn – host

So we were talking about your experience in the packaging industry. Sounds like primarily, right? Where in there did machine vision come in? How did you end up making that move?

Armando Bompane – ATEK

Yeah, so in fact, there was a project that I got involved in where the OEM had sold some robotics that was integrated with machine vision. It was quite a complex robotic system that was processing some food products. And there was some significant challenges in that project actually. And before that, I hadn’t been doing too much in machine vision. I’d be more focused on the automation and other parts of the integration. And that project had fallen on my lap and I had to sort of deep dive into the world of machine vision without having done too much before that, but having a lot of automation experience and manufacturing experience and trying to unpick, okay, what are the challenges in this system and how do we actually make this work to the customer’s requirements? And it turns out that what happened in that project was that there was a mismatch between customer expectation versus what the OEM could actually deliver for that product, which is quite, quite common across the board. And so my, my goal was helping the customer define actually what is the requirement. So actually, what are you trying to measure? What is your quality specification?

Armando Bompane – ATEK

How are you measuring it? What is a defect? What is a pass/fail look like? And then translating that into some information that a machine, that the machine vision system could interpret and make decisions from. So I came in at that interface layer between OEM and end user and bridging that gap. And actually, if I really summarize ATEK in one line, it’s really, that’s where we sit. We sit between the technology and the end user and we help bridge that gap. We help the end user understand how to digitize the information they currently have off their production line., and then we develop the software, the machine vision software, to deliver that solution for the, for the end customer. And it’s a fascinating space to be in because we’re actually developing machine vision algorithms and then we’re integrating the algorithms with the physical world around us. We’re integrating with automation and robotics and things like that. Hey, Mr. Customer, you want to automate quality inspection. What does that actually mean for you? If we over, if we make this threshold too sensitive and you get too many false positives or false negatives, what does that mean for your production? Have you thought about this? What does a false positive, false negative look like? How do we define that so that we get the right metric so that you can still produce efficiently and to your customer requirements without causing any unwanted stoppages of your production line? A lot of our work is actually spent in the customer domain around translating their subjective quality information into machine-interpretable data. That’s basically a lot of what we do. And then in the backend, we could probably say under the hood, what we’re doing is we’re deep in the, deep in the algorithms. We’re basically creating custom tool sets, which is a combination of machine vision algorithms to solve these problems. And these problems could be anything from a measurement problem. You know, you want to take the measurement of the neck on a bottle. And you want to map standard deviation of measurements and understand what is your accuracy of your molding machine, or you want to make sure that your food product is labeled correctly. You’ve got potential allergens that could be in the product, and you want to make sure that end users are aware of that. So you want to make sure the right label is on the right product.

Armando Bompane – ATEK

You want to make sure that hard-earned dollars that mums and dads are spending now at the supermarkets— you bring home an expensive block of cheese, you want to make sure it doesn’t spoil early. So how do you make sure that that seal is intact before it leaves the production line? And then how do you, how do you actually measure that? And it turns out that when you have this conversation with, with the end user who, who doesn’t work in the machine vision space, they’re not thinking every day in algorithms and in, in mathematical terms about computation. They’re thinking about how do I get my product out the door and make my customer happy. It turns out when you have that conversation with the customer they go through this aha moment and they start thinking about, okay, what does quality actually look like? And what I try and— the journey I try and take customers on is if we focus on the quality of the product and how we measure that, the insights we get from quality affect all of your production. They affect efficiency, they affect reliability, they affect downtime. Because there are, there are parameters in the quality of the product that we can detect that can make, that can allow us make an inference or some kind of prediction around there’s an issue in this part of the production line potentially that you need to resolve.

Armando Bompane – ATEK

So there’s some really interesting and powerful insights we find in the data that we collect from, from the quality side that has not only benefits in, in quality specifications, but around the total production line itself.

Josh Eastburn – host

Yeah, I think I’m, I’m really loving this because I’m imagining all the different kinds of listeners, right, who listen into, to the podcast and how this could be useful for them, right? Maybe it’s a system integrator who’s, who’s not as far ahead in their journey as you are, but it could also be an end user who is like, I’m struggling to get people to understand what it is that I’m trying to accomplish here, the problems that I’m trying to solve. In your experience in having this conversation with lots of different end users, where does that communication tend to break down or tend to be the most difficult? Why is it that they’re struggling? You mentioned obviously they’re not living in vision algorithms, they’re not used to thinking in those terms. But yeah, what have you learned about maybe making that process easier?

Armando Bompane – ATEK

I think the biggest gap in understanding or yeah, in understanding by end users and just general consumers is that they think that machine vision is, is really about a camera. Sometimes I get asked the question, if I buy this camera, can it do XYZ? And I think fundamentally why machine vision is a challenging field is because the end user sees it as a camera, but machine vision specialists understand that there’s this whole hardware-software stack involved. You’ve got photons of light that are projecting off an object, and you need to make sure that they’re projecting off in a certain angle. And then you’re capturing that off an image sensor through a lens. What kind of lens do you select? What kind of image sensor? What kind of light? How do you angle the light? How do you orient that on a machine? And then how do you put together the toolset of algorithms to make a decision, right? And that is fundamentally what the machine vision integrators are working through, this complex technical stack between hardware and software interface. And on top of that, you’ve got challenges with network, with bandwidth and things like this. You’re trying to make this whole system with multiple layers of hardware and software.

Armando Bompane – ATEK

You’re trying to convert that into a deterministic package that can run on a production line 24/7 without failure and make decisions for the customer. And fundamentally, there is this gap between what end users see as machine vision and what machine vision integrators know machine vision is. And our job is helping bridge that gap and making sure that end users understand there are really amazing possibilities of what you can do with machine vision, but there are also constraints and limitations as well. And how do you, how do you manage that to make sure the system is actually delivering long-term value for the next 5 years? And I have this conversation a lot, and I’d love for people to give me feedback on my mental model of the market, but I, if I was to map the whole machine vision market as a bell curve right? So let’s just think of a normal distribution bell curve, right? And the x-axis is complexity and the y-axis is the number of potential users, right? So if I was to map the whole global market like this, machine vision and manufacturing, I think the mean of the bell curve, which is where the peak is, is your general applications.

Armando Bompane – ATEK

So very, very general standard applications. So is my use-by date printed correctly? Is my label on correctly? Is my cap on the bottle? These kind of generalized sort of applications is the mean of, of where all the applications sit. And then as you move out in, as you move across a few standard deviations either way of the bell curve, your complexity increases. And that middle part of the market where most of the applications sit is what I call the smart camera market. So this is where the smart cameras are packaging their solutions into a camera that is maybe an edge-based camera that can do a few bits and that can do a few things. Let’s sprinkle some AI in that as well, and that’ll solve a lot of these general, these general problems. Okay. And then you have all of these, but that is like what you find a lot in this part of the market is that customers will go and buy like a smart camera, and smart cameras definitely serve a purpose. They solve a lot of applications well, and we actually are productizing something in a smart camera format, so this is nothing about smart cameras specifically, but what you generally find in that smart camera market space is that customers buy it to solve a problem today, and in 12 months’ time, the production line changes.

Josh Eastburn – host

Hmm.

Armando Bompane – ATEK

The process changes, and the customer is like, how do I retune this? Like, I’ve only got this shape matching basic algorithm, and all it’s doing is finding the ice cream ball on the label. It’s not telling me what color the ice cream is so I make sure it’s the right colored ice cream label. That’s a real use case, by the way, that I had a conversation on.

Josh Eastburn – host

I believe it.

Armando Bompane – ATEK

Right. And it turns out that a lot of these cameras end up either being underutilized or turned off, and customers don’t get the full value of that smart camera over the lifecycle. So what we’re doing and where we see the market heading is all these complex projects that are not smart cameras, they tend to be very project-based. So they tend to be like, we’re not talking $20,000 smart cameras, we’re talking $200,000 projects potentially that are multidisciplinary. There’s an R&D phase involved and then you’re validating and then you’re designing and integrating things like that. We’ve been operating a lot in that very complex technical project space and where I see the market heading towards is that someone needs to fill the gap. Between the smart camera or the usefulness or the simpleness of a smart camera, if that’s such a term, and the complexity of what we learn in the project. How do we package that into a format that gives the customer long-term value so they have a very low TCO, low total cost of ownership, and a very, very high long-term value and utilization of that asset? And what we’re trying to build is we’re trying to build a platform platform that combines the two.

Armando Bompane – ATEK

It’s how do we, how do we package something in a simple smart camera format, but how do we allow the end user to be able to customize and modify without an expensive integrator coming in down the track and making modifications? And a lot of the work we’re doing is around taking these lessons and trying to package that in a format, which we’re hoping to do over the next sort of 6 to 12 months with a product release.

Josh Eastburn – host

When you talk about packaging that in a format, is that about the user interface, making it easier to, to retask the device? Like you said before, maybe they can do it for the first project, but they can’t do it for the second project. They just don’t know how to work with it. Or what is the problem that you, that you really think is at the heart of that? If AI can’t just solve this on its own, it’s not smart enough. There’s some user component there that’s required.

Armando Bompane – ATEK

Yeah, I, I think the AI conversation is like, that’s probably a podcast in itself, as it’s a huge topic. And I think we see definitely a lot of benefits with AI used in the machine vision space. Absolutely, AI and particularly some of the latest anomaly detection models are fantastic at solving generalized defects and allowing you to retrain and retune as time goes on. Do we need to use that technology all the time? You know, like, is it sometimes a bit overkill using that technology? AI is as good as the data going into the system. You know, we all know fundamentally the AI is some kind of probabilistic model, and the quality of the data going in is going to map somehow to the, the quality of the data that’s, that’s coming out. So I think fundamentally in the, in the AI space, yes, there’s, there’s some great use cases for it for generalized defects and for ongoing training, but I think there’s definitely going to still be an ongoing need for traditional rules-based machine vision algorithms. And I think the challenge is around when you deploy a traditional rules-based system, how do you make that retunable and trainable in 12 months’ time and, and in 2 years’ time to an end user without giving them a simple interface to upload new images and train a new AI model, so to speak.

Armando Bompane – ATEK

And I think this is where the power of some of the new LLMs are coming into play. So, you know, LLM interface, large language model interface to the vision algorithm toolset, which then maps to the output from that machine vision system. We’re doing a little bit of work around the LLM interface into that. And maybe that gives the customer or the end user who’s not necessarily a programmer, they’re not necessarily an engineer, but they can explain in natural language the defect to a degree that allows a large language model to convert that into something. That then a machine vision model can run with. And we’re dabbling in that space at the moment. We’re already seeing that in some machine vision systems that are out there. People are using this kind of interface. And I think that’s definitely going to become a larger part of what machine vision companies start offering over the next couple of years. I think these latest vision models that are even coming from the large AI companies are really powerful. I mean, there’s some amazing models around You know, like, let’s take the machine vision problem of depth, which used to be a stereo vision problem, and now estimating depth from just a standard 2D image, you know.

Armando Bompane – ATEK

And some of those models that are doing that are quite impressive. So that automatically simplifies the hardware stack in doing that. But I think fundamentally the physics problem will always be there, and the physics problem is always going to be particularly with something that you need accurate measurements from, or you need some, some level of accuracy from. It’s not just a generalized, these sort of defects, or these, these dents on a bottle, and we want to, we want to generalize what these dents look like so that when we produce these bottles on a production line, we don’t have any, any dents. So I think there’s always going to be that space, but I think fundamentally there’s always going to be a space in machine vision where we need a deterministic system that can make these decisions. And fundamentally, that physics problem is always going to be there, I believe. And that’s the space we play in. We play in the space of solving the physics problem at an engineering layer and then having our software sit over the top of that and packaging that for an end user to use. That’s the space we play in every day.

Josh Eastburn – host

Mm-hmm. So it sounds like the problem that you’re trying to solve with the platform you’re putting together is partly a usability problem and maybe also a comprehensibility problem. Like we’re not working with a complicated deep learning algorithm that’s hard for the end user to understand. We’re working with something that has hard-coded rules or intelligible rules that could be explained to the person who’s trying to operate the system. Is that kind of what you’re explaining?

Armando Bompane – ATEK

I think for sure the market will need to head in this direction in order for the market to continue growing. We have a lot of these, as I was referring to that bell curve before. There are a lot of applications that are currently sitting outside of the smart camera product market space, right? And those applications are extremely complex, and a lot of them require R&D phases, which we get involved in with clients. We do the R&D, the validation, and then we build the system from there. But I think the market over time, a lot of those previous projects that used to be very technical and complex and required R&D AI is actually solving some of those or doing the heavy lifting for some of that work and allowing us to map those complex projects into a more standardized space. And I think AI will help us with standardizing some of this complexity and packaging it in a format that, yeah, maybe this complexity is packaged into a smart camera and I can bolt it onto my production line and keep making informed decisions and, and get value out of it for many years. But I don’t think the market is there yet from what I can see.

Armando Bompane – ATEK

But I think it’s definitely heading in that direction. I mean, I’m— you would see it in your role with, with everyone that you’re talking to around the world. There’s just every day there’s new products. Everyone is working on something. And I think, I think essentially machine vision is this, is untapped market, global market. I think it’s, it’s been previously reserved for specialists that have maybe come out of university and done PhDs in machine vision algorithms. And it’s been reserved maybe for camera companies and things like that. And I think fundamentally what happened in software 20, 30 years ago when things were being open sourced and everything exploded, I think machine vision is moving in that direction where it’s a huge untapped market. And people think the only data you can get out of machine vision is something in reference to, to quality of a product, when in actual fact you can get a lot more powerful data. I had a conversation with a client last week, really fascinating conversation. This was in a meat processing plant, right, an abattoir, okay? And it was a safety-based conversation. They had a safety incident in the plant., and we were walking around the plant and we were talking about, you know, we were talking about quality, but then the conversation moved into safety and this discussion about this safety incident.

Armando Bompane – ATEK

And they were talking about how do we map out an exclusion zone for the operator and do we have to build a fence here because the operator should not be standing in this place, it’s a dangerous space. And I said, look, well, you’ve got a video camera up in the corner, why don’t you just plug the AI into the video camera and map an exclusion zone and use that as your detection for the exclusion, and sound an alarm. And the customer said to me something really interesting. If I use the AI from the, from the video feed for a safety exclusion zone, can I give it a SIL rating, which is like a safety integrity level rating?

Josh Eastburn – host

Yeah.

Armando Bompane – ATEK

And I paused for that moment and I thought about it and I said, huge question. That’s a huge conversation. Let’s just take that offline because we can’t We can’t answer that walking through a production floor. But I went away and started thinking about that problem. Well, hang on, how do you give a probabilistic model a safety integrity rating, which in manufacturing, as being an automation engineer’s background, you need a deterministic system to do that. That’s why PLCs are so strong in these environments. And then I started thinking, what if it is stereo vision So if I use stereo vision, so two cameras on top of the exclusion zone looking at people walking through that exclusion, and I’m mapping depth perception, and I’m using standard machine vision to solve that problem, can I give that system an ASIL rating? All I’m doing is mapping contrast of pixels. I’m not putting it through a probabilistic model. Then I got thinking, what if I used optical flow as an algorithm and I was just tracking these contrasts of pixels across a certain region of interest in that, in that space? Can I give that a safety integrity level rating?

Armando Bompane – ATEK

And this is where I think the conversation between traditional rules-based machine vision and AI is quite interesting. So the problem statement, how do I give a system a safety integrity level rating? Does that system need to be deterministic or can it be probabilistic? Now there’s an argument in the AI space around, well, humans are basically probabilistic by nature. In how they make decisions. So how do you give a safety officer who’s policing that area, how do you say they’re deterministic? Okay, well, that’s that argument there, but fundamentally, that it’s still, it’s still a computer system, still, still an AI model that’s making decisions. And then it got me thinking, well, traditional machine vision rules will probably fundamentally always have a place, and I think particularly in that safety space as well. I think that’s, I think that’s a fascinating, fascinating question. I’ve sort of digressed a little bit, I think maybe from the initial point, but it’s a fascinating question that I think is important for us to unpack.

Josh Eastburn – host

Yeah, that is actually really insightful. I’m really glad that you brought that up because yeah, the core problem of AI where it is today is explainability, right? And you have folks who are just focused on solving that one problem of help us understand what this black box is thinking, why did it make this decision before I trust it? With my business enterprise or my health or something like that, right? Not a solved problem. So this is actually, you’re actually tackling that in, yeah, in maybe a counterintuitive way, right? I think most of us are thinking, well, to solve AI, we probably need more AI. And you’re actually saying, hold on, we’ve kind of solved this problem before, right? We know how to codify, we call them algorithms, right? We know how to codify processes and we call them algorithms and those can be explained. So that’s actually really interesting. I really appreciate that.

Armando Bompane – ATEK

So I’m working on a machine vision problem at the moment. I have this part and the customer has potentially all these different defects happening on the part that you just can’t use traditional rules-based systems for because this deformation or this defect in one day might look like this and another day might look like that. And to put a rule around that is nearly impossible. You’re not going to have a system that’s, that’s going to have long-term reliability. So the flip side there is Do we use deep learning to solve that problem? Okay, well, let’s go get a bunch of images. Let’s go train anomaly model and then let’s deploy and then let’s retrain and retrain and retrain from there and optimize, you know, or do we need to sometimes necessarily do that? Can we use a combination of both worlds? Can we use a standard off-the-shelf AI model, pre-trained AI model, and can we use just the inference of that model for feature extraction and then can we compare the features of a golden image versus the features of a defective image and use that for image classification or for defect identification. And it turns out that for a lot of use cases, we find that using a little bit of both actually is very, very powerful and it allows very, very fast-to-market times to deploy because we don’t need to go and collect this massive dataset to do training, which sometimes has its challenges in itself.

Armando Bompane – ATEK

Sure., and then sometimes you need to go into synthetic data and that’s a whole different, whole other level as well. But sometimes we can use the best of both worlds to solve a problem very, very quickly and deploy very, very fast. And being able to sit objectively like we do in the space and look at all technologies objectively and not sort of jump in and say, hey, AI can solve everything, but objectively looking at it and say, let’s use some parts of AI, let’s use some parts of a traditional-based system. And then let’s put them together and let’s use that to solve a problem actually sometimes means we get better outcomes for clients.

Josh Eastburn – host

Very interesting. Is there a chance, or are you hoping to bring something to market in 2026?

Armando Bompane – ATEK

Yeah, so speaking of the smart cameras before, we are thinking about packaging some of the complexity of the learnings of these real complex projects into something that is an edge-based device, because I do think fundamentally something that is very simple to deploy and integrate ultimately becomes easier for an end user to use, and then they can get more value by reusing that device over multiple production lines or moving it as time goes on. So I think for us, a lot of our focus is around packaging complexity into an edge-based device, which the user interface will be a served web page where they can log into it and they can They can look at it on their tablet or something like that. A very, very easy to deploy device that also gives them the freedom and complexity to be able to solve more complex problems. So it won’t be a smart camera. It’s not going to be something that will have a generalized user, a generalized training system. It’ll be something where we’re still heavily involved as the integrator, helping to do retraining and upgrade the system for more complex problems., but we’re going to try and package that in a way that customers can deploy very, very quickly and easily into production lines.

Josh Eastburn – host

If somebody is interested in what we’ve been talking about here and wants to follow along or keep up with announcements about what you’re working on, what’s the best way to do that?

Armando Bompane – ATEK

Well, we do have a website, atekintengineering.com.au, And you can also find me on LinkedIn and you can find ATEK on LinkedIn. So we post a lot of content on LinkedIn about the applications that we’re working on. And we really try and just communicate the value of those applications to our network. So please follow us there and reach out to us for a conversation. We love understanding our customers’ problems because we learn so much from these conversations and they’re so enriching to have. What I get the most enjoyment of is working with a customer and taking them through that journey of translating quality specifications into something that machine vision can solve. And then talking to that customer in 6 months’ time or 12 months’ time and seeing them still get good value out of that system. And I get real enjoyment out of like even years down the track and customers have run tens of millions of cycles through the system. It’s like, hey, we haven’t heard from you for ages. How are things going? No, things are good. Things are really good. It’s, it’s, everything’s great.

Josh Eastburn – host

Good news.

Armando Bompane – ATEK

You know, so please reach out to us. We love talking to you and, and learning more about your challenges.

Josh Eastburn – host

That was Armando Bompane, founder and engineering solutions director at ATEK Integrated Engineering. The core idea I’m taking from this conversation is one worth sitting with. The reason so many machine vision projects underdeliver has more to do with the work of translating subjective human judgment into objective, machine-readable criteria and foundational engineering design that is made to last. Get that foundation wrong and no algorithm is going to save you. If you want to learn more about ATEK’s work in machine vision, intelligent inspection, Automation for Manufacturing, visit them at ATEKINTengineering.com.au or find Armando on LinkedIn. If you’re a vision professional or an integrator with a project story worth sharing, I would love to hear from you. Find me on LinkedIn or reach out to josh.eastburn@mvpromedia.com. This episode was produced by Flannur Creative Studio and Big Robo. For MVProMedia, I’m Josh Eastburn. Be well.