TechCasts

Read the Transcript

Ben: Hello hello, welcome back everyone to our SYSGO Techast where we share insights from the thrilling world of embedded applications and talk about our products, company and services. I am Ben from the Marketing & Communications team and in this episode I am accompanied by Joe Richmond-Knight. Joe is member of our FAEs and we are going to hear more about his technical background and what great projects he supports. Have your tea ready? Then let's start. Welcome Joe.

Joe: Thank you Ben. My name is Joe Richmond-Knight, I am a Field Applications Engineer for SYSGO. I have been with the company for nearly two years now, I think. Based permanently in the UK but work globally with the rest of the technical Sales and the Sales team. Previous roles, mostly been in IoT, so Internet of Things, Edge compute systems, both on the Sales side but also the operations side and the engineering side, so I really managed to get quite a good amount of experience in a lot of competencies. So yeah, great to be here.

Ben: We are delighted to have you here. Now diving a bit deeper into your journey, we are curious to know how did your first come across SYSGO?

Joe: I was working for a technology company at the time and was actually contacted by one of our colleagues here in the Sales team. And we had a mutual friend, so it just goes to show the power of networking really. He was explaining what SYSGO did, the core competencies, our products and the kind of industries that we work in. And I thought this sounds like a really interesting challenge.

It's the first time working in the technical Sales team is actually the first time of, I suppose, fully acknowledged a Sales interest in myself. I've always been, I would say, an engineer or logistics-oriented guy, but I would say maybe secret Sales. I would work with a Sales team and another organization helping with solutions architecture, scoping of systems, etc. And frankly, I actually didn't know that this kind of role, so a solutions architect or a field applications engineer, was the role. So when Mark and I had a number of conversations, I actually thought this is perfect. This is exactly what I wanted to do. I was really, really happy when the opportunity came up.

Ben: So you feel happy here and you love what you do.

Joe: I think the breadth of technology that SYSGO works within, so the mixture of hypervisor technologies, real-time technologies, different operating systems is really interesting. So from a technical standpoint, I was quite gripped. From an industry standpoint, I've always had a keen interest in railway technologies and aviation and avionics, because I think, frankly, everybody who works for SYSGO is a bit of a geek in the nicest way possible. So it's been a really nice fit, both from a technology and the kind of organizations we work with as well.

Ben: So would you say there's a difference between technology and, in our case, embedded technology?

Joe: I think there's a big push at the moment for the SWAP, so size, weight and power, challenges in embedded technologies. So I think certainly to work in this kind of space where these kind of systems are used presents unique challenges. I suppose I was quite lucky in the sense that working in IoT, you know, the Internet of Things, where a lot of these devices are very small and require or consume very small amounts of energy. There was a decent amount of understanding there already.

I think one thing that is different, of course, in the areas that we work is the cost of these systems. I think, although they are still embedded products because they are safety and security. Otherwise, something you would consider a mission-critical system, that's a bit of a change. But again, it's interesting to work with expensive and cool technology.

Ben: That's brilliant. Since you've already done so much, do you see any future technology trends?

Joe: I mean, of course, it's a difficult question to answer if you look at every industry, in all industries in their totality. I think every industry has specific technologies that may or may not be more important in the future. But I think, of course, the topic of artificial intelligence right now is quite interesting. It's a hot topic, both in a good way and sometimes in a bad way. But in the embedded space, in our space, I think there are some really interesting challenges that we can solve both with embedded AI and Edge-based AI solutions.

So, for example, I know we're seeing a lot of interest in the Automotive world for predictive maintenance for vehicles. So instead of waiting for the vehicle to break down or simply relying on a scheduled interval, whether there may be a problem or not, I think you can leverage some of the AI capabilities that are coming out in modern chipsets, in modern SoCs, to really leverage some of the data that's coming from these vehicle platforms.

I know in the past, lots of vehicle manufacturers send to military up to some kind of central server, but there's an interesting use case for putting that technology in the edge, in the physical device itself. Of course, you will always have a limitation on processing power when you have these systems in the Edge. So, the centralized, or I suppose you would call it more traditional model of artificial intelligence in the cloud, still makes sense. But of course, if you have a mixed criticality, either in the terms of safety or security system that you're designing, you still need a safe and secure method of transferring that data up to the cloud as well.

It's been really interesting to work with a number of our customers as to what that looks like in the context of PikeOS, PikeOS for MPU, ELinOS, our core competencies. 

Ben: Yeah, we see similar use cases like scanning a canal for damage or cracks with a drone, or AI image analysis for train wheel maintenance and so on. So embedded AI is really on the rise!

Joe: I think that's a really interesting use case because, of course, there are use cases for machine learning, AI platforms in an offline system, whether you consider a modern car on an online or an offline system nowadays is a separate conversation. But actually a problem that we still face to which we have been facing for a number of years is the cost of the internet. It's the cost of broadband has gone down. However, when you're looking at sending gigabytes and gigabytes of information over a wireless network, so a cellular network, that's still relatively expensive. And to leverage the kind of benefits of AI, those systems need a lot of data.

So in a traditional model where you would say have a relatively "dumb" device on the Edge, you know, in some pictures, sending those pictures up to a cloud server and every single one of those is being analyzed by the system. Yes, the architecture is simple, but actually in reality you're wasting 90% of your valuable bandwidth on sending junk data. So by putting that intelligence in the Edge in the device, which it actually contains the sensor, you can be really efficient on data usage. And so it's very important for Edge-based AI, extremely important, even in the case of low cost IoT systems.

For example, a popular application and one that I have a reasonable experience with is predictive maintenance and risk management with IoT systems. 99.9% of the time, you hope, if you are monitoring some sort of asset, there isn't a problem. It's that 1% where it really matters. Therefore, it's very wasteful to send the exact same data 99.9% of the time, with nothing changes. So if you can implement these models in the Edge using these various AI chipsets and AI tools, then you can be very cost effective on the way that you transfer that data. 

Ben: So it's all about having intelligent models already in your platform, like in your hardware.

Joe: Exactly. You just need enough intelligence in the Edge to be able to make a decision on whether the data is valuable, not necessarily what the data actually means. And I think we're seeing a lot of applications where Edge technologies are used for the exact reason. Is the data valuable? Is the data junk? Yes, no. If data is valuable - send. And then do more analysis on the data with a more performant and a less restricted system. So it's almost a two-stage approach.

Ben: I suppose it's all about filtering the data, isn't it?

Joe: Understanding which data is valuable, which data is less valuable. And when you have limited resources, whether those resources are people, or whether they're computing platforms, making sure that they're acting on data that they really should be acting on.

Ben: That topic is really fascinating, and there's so much more to talk about. In part two, we will dive deeper into our products and solutions. But that is it for today. Thank you for tuning in and connect with us on our social media channels, wishing you back three days and glitch three nights until we meet again.

Goodbye from the SYSGO Central.