How can you use digital twins and simulation to meet KPI goals and show ROI? Sam George, Board Member of Cosmo Tech and formerly Corporate VP of Azure IoT at Microsoft, joins Ryan Chacon on the IoT For All Podcast to discuss digital twins and simulation in IoT. They explore what digital twins are, the value of digital twins for KPI and ROI, simulation, the role of AI in digital twins and simulation, and the future of digital twins and simulation.
Sam George is a board member of Cosmo Tech, a digital simulation solutions company. Before Cosmo Tech, Sam was Corporate Vice President of Azure IoT at Microsoft, responsible for Azure IoT Central, Azure Digital Twins, Azure IoT Hub, Azure IoT Edge, Azure IoT Hub Device Provisioning Service, Azure Maps, Azure Time Series Insights, Industrial IoT & Manufacturing, Automotive / Microsoft Connected Vehicle Platform, Smart Buildings, Smart Energy. Additional responsibilities included IoT business strategy and results for IoT in Azure, IoT product offerings and roadmap, developer facing features, integration with other Azure services, partner and field enablement and more.
Sam spent over 20 years with Microsoft and has spent time in all three engineering disciplines (development, test, and program management). He spent most of his career as a dev manager, dev lead, or developer and in 2011, he switched to a product leadership position.
Interested in connecting with Sam? Reach out on LinkedIn!
About Cosmo Tech
Cosmo Tech is a French company based in digital twin simulation solutions for international and complex enterprises.
Key Questions and Topics from this Episode:
(01:39) What are digital twins?
(05:39) Where simulation fits in
(11:27) When not to use simulation
(19:36) Learn more and follow up
– [Ryan] Hello everyone and welcome to another episode of the IoT For All Podcast. I’m Ryan Chacon, and on today’s episode we’re going to talk about digital twins as well as what simulation is and how it plays a role in the Internet of Things. With me today is Sam George, the board member for Cosmo Tech. They are a digital twin simulation solution provider for international and complex enterprises.
In the past, something being really interesting I wanted to mention before we jump into this is Sam was the corporate vice president for Azure IoT. Obviously, all of you know how big of a role Azure plays in the IoT space, so his insights are super fascinating, and I think you’ll get a lot of value out of this conversation.
Give this video a thumbs up. Subscribe to our channel if you have not done so already and hit that bell icon so you get the latest episodes as soon as they are out. Other than that, let’s get onto the episode. Welcome Sam to the IoT for All Podcast. Thanks for being here this week.
– [Sam] Hey Ryan. Good to see you.
– [Ryan] Yeah, great to have you.
Let’s kick this off by having you give a quick introduction about yourself, your background experience, information about the company you’re with, to our audience. That’d be great.
– [Sam] I retired from a 25 year career in Microsoft. In the last eight years of those, I started and then led the Azure IoT team at Microsoft. Today I’m working with a couple startups. One of those is Cosmo Tech who does AI guided simulations around digital twin mapped environments.
So that’s what I’ve been up to.
– [Ryan] Fantastic. I remember when I was first introduced to you and having you on the podcast, I was very excited, not just because of the Azure background and obviously how that fits into the IoT space, but what you’re doing now, is quite relevant. Digital twins, talking about simulation.
I’m very excited to dive into that a bit further today. But maybe what we can do is just have you quickly just describe to our audience who may be a little bit less familiar with some of these terms, what exactly digital twins are, how they work, and the value they provide for the industry.
– [Sam] Sounds great. If you think about start with IoT. The whole reason why companies take advantage of IoT is to sense what’s happening in the physical world in real time, right? Whereas before you didn’t know, you put a sensor in the physical world, you can know in real time.
What’s interesting though is the more and more IoT sensors that you get in the world, you start to need more context about where they are and what structure they’re a part of. I’ll give you an example that most people I think would be pretty familiar with. If I’ve got a smart building and I’m trying to monitor things like occupancy, it’s not that I put a thousand sensors in a skyscraper and just call it a day, right?
I actually need to build a model of the skyscraper, like a virtual model. Rooms, floors, elevators, things like that. And then put all the different sensors in that context. So an easy way to think about digital twins is it’s a digital replica of a physical object, but that digital replica then also provides context for all the IoT devices that are reporting as part of it.
So if an occupancy sensor is starting to move in a smart building, oh, that’s on the fourth floor in room six as an example.
– [Ryan] Yeah, that’s great. We talked about digital twins a good bit, but I always love starting with an introduction and high level overview just to have somebody else explain it in a different way. And I think, it, what resonates with our audience obviously varies, but when it comes to digital twins being used in these situations for these use cases that you’re talking about, how are they really helping drive and show ROI, meet KPI goals, these kinds of things outside of not just what they are, but what are they really doing from a value standpoint as it relates to those areas?
– [Sam] Well, one of the first things you can think, you can almost think about a journey that a customer goes on. One of the first things that digital twins is good for is operational awareness. And again, take that smart building example, or take the example of a factory or a construction site or, any number of things that you could use to map the physical world.
Step one is to be aware of what’s going on. Once you’re aware of what’s going on, it leads to those questions of why is this happening? The other thing that operational awareness gives you is over time, it gives you a corpus of historical data and then that historical data can be used to find trends and anomalies and things like that, and then to project forward with patterns.
And so operational awareness is the first thing that you can do with digital twins. But companies like Cosmo Tech that I’m working with, they’re going way beyond that, and they get into something that’s related to simulation digital twins. And so what simulation digital twins is it takes that operational awareness and it says, let’s project forward. Not only all of the possible states of this system into the future, but let’s guide the system towards, what we wanna achieve there.
So for example, in a smart building, if I wanna reduce the amount of energy that I’m using in a building, or if I’m building my next factory and I want to figure out the best way, the best place to locate it so that I reduce carbon emissions and have supply chain resilience, simulation digital twins can take something that you know about and figure out where you want to go, and then map out all of the millions of possibilities of how to get there.
– [Ryan] And so when you mentioned simulation, it’s interesting because when I got started in the space about seven years ago now, simulation was a topic that was brought up back then a good bit. This was almost before at least digital twins was heavily pushed mainstream kind of topic and the value simulation provided.
But then it felt to me at least, that conversation around the value of simulation or the implementation of simulations died off a bit and now it’s coming back. And can you just talk more as it relates directly to simulation why or how it plays into all of this.
And you’ve obviously touched on what it is, but just how does this now playing a role in driving support and success with these implementations when it comes to IoT solutions and just dive in there a bit further.
– [Sam] So, couple things. First is that when most people think about simulation, they think about a 3D environment that looks like a video game or something like that. And that is part of simulation to a degree, but you know that, that’s not really the interesting part. Some of the examples that you see in the industry are, Hey, I’ve got a factory and I want to use 3D simulation to figure out where to place all of my equipment before I ever start placing equipment.
That’s one type of simulation. The simulation that I’m talking about and that Cosmo Tech does is remarkably different. So a good way to start explaining how this type of simulation works is to think your own personal experience as a consumer using maps, using something like Google Maps as an example.
Let’s say that you’re going to a location that you’re not familiar with. You put in the location, you put in a starting point, which is typically right where you are, and then you get directions, right? What you’re not seeing is under the covers, what systems like Google and others are doing is they’re mapping out all of the thousands of possible routes
to get you there. They’re considering all sorts of obstacles that are in the way. There might be construction at this place, this road might be closed. And then they’re showing you an optimal route, right? And then you start following that and along the way it gives you turn by turn directions.
And what happens is if you miss a turn, of course it’ll reroute you. And all it’s doing is it’s figuring out you didn’t take that route, so I’ll give you the next best one. Now imagine if I am someone in this example like Cosmo Tech works with a company, Michelin, that builds tires.
And I wanna do something super challenging, which is I want to figure out the optimal place to source new production to serve the Chinese region. And I wanna do it based on a whole bunch of KPIs. Those KPIs include things like, I want to be as sensitive to carbon as possible and reduce carbon. I want to increase production, I want to increase profitability.
I want to reduce times through customs. Now that’s a really, the analogy there is, Michelin is effectively wanting to put a pin on a map and say, how do I get there? That’s a whole different problem. And what companies like Cosmo Tech do with simulation is they can simulate all of the different possible places to put a factory and consider all of those different KPIs.
How long is my weight gonna be through customs as I’m shipping tires, how much carbon am I gonna be using? And then it figures out the optimal way to get to my KPIs. And so as it relates to IoT projects, that’s an example of I’m building a new factory somewhere.
As it relates to IoT projects, one of the things that goal seeking simulation can do is it can figure out is this even gonna be a viable IoT project? So I can avoid getting stuck in purgatory. What are the benefits that I could hope to achieve with this solution? And what is the cost going to be of that solution?
And is the ROI right or isn’t it? And so simulation in that case, where I’m projecting forward, can really save me a failed project or in the case of Michelin, it can save me increased profitability by 5% and reduced carbon emissions in this case by 60 percent.
– [Ryan] How early on in the process does simulation really come in? It sounds like it’s early on in the state, in the process of, once you have an idea of what problem you’re trying to solve, you’re starting to put the pieces together. Before you really invest the time into the physical elements that usually have a higher cost to change and fix you would probably bring simulation or it sounds like you would bring simulation in then in order to project different pieces of it in order to save costs, save time, make the right decisions as best as you possibly can.
So can you talk us through how you would approach incorporating simulation into that IoT journey a company would go through you know, what’s needed, what needs to be done, how does that all work?
– [Sam] Well, one of the first things that you start with before you undertake any IoT project or anything related to IoT or digital twins is what’s your goal? What are you trying to do? What are you trying to make better? What are you trying to make more efficient? What are you trying to reduce? When you have that in mind, then you know, if it’s a simple IoT project and I’m doing something like, monitoring footfall traffic or something where I don’t need digital twins, that’s really just a sensor-based, project where I’m looking for operational awareness.
But in cases where I want to do something more remarkable, in cases where I’m gonna go take advantage of mapping an environment with digital twins, once I know what I wanna do, like in the case of Michelin, once I know what I wanna do, which is increase profitability, reduce carbon, then the first thing that I do is build a digital twin model of the factory, the supply chain, the constraints and things like that.
And then you’re absolutely right. Before any plant gets built or sensors get laid or anything like that, it’s like, go and simulate the whole thing, right? Take advantage of companies like Cosmo Tech, which provide that goal seeking simulation that says, what is the maximum I can hope to achieve in this case?
Think of it as step one, have a great idea of what you want to do. Step two, model the whole environment. Step three take advantage of goal seeking simulation in order to achieve your KPIs and see whether it’s possible.
– [Ryan] Are there any use cases or situations where maybe simulation isn’t necessarily a fit. It sounds like it’s gonna be a fit for pretty much anything you’re venturing into when you’re bringing an IoT solution to life. But do you ever find a scenario where simulation just might not be appropriate?
– [Sam] I think it’s a great question and the way I think about this is, the simpler the solution is or the simpler what you’re trying to do is, the less need you have for simulation, the more complex the environment. If you’re talking about global supply chains, or I’m talking, you’re talking about a really complex environment where even the experts in that environment have trouble keeping everything in their heads.
That’s where simulation really shines, because simulation takes the power of software and it can simulate millions of possible decisions. In the case I brought it up with Michelin. In the case of Michelin, if you think about, I’m sourcing new production somewhere, how do you typically do that? What you do is you have experts that have done it before and they take everything that they know that’s in their head.
They take their best guess and they say, I think it should be about here. And sometimes they’re right and sometimes they’re not. What goal, AI guided, goal seeking simulation really does is it gives those expert superpowers, right? It’s really there to help them and assist them in taking that complex environment that’s really difficult for a human being to consider all possibilities and effectively generating all the different possibilities and coming back with, here’s the optimal one.
So more complex a scenario, the more use you have for simulation.
– [Ryan] Perfect way to put it. I like that. You mentioned AI, I wanna actually dive into that for a second. So talk about AI and digital twins and how they play a role together when it comes to simulation. What’s being done there? What’s the value? How does that all work?
– [Sam] Well, In the case of Cosmo Tech with goal seeking simulation, the reason why AI plays an important role is because AI is the thing that helps refine and guide the simulations towards the KPIs. If you think about a really naive system that’s doing simulation, what it would do is it would just do some brute force permutation of every possible location to source a factory on the planet.
You could maybe take a billion different possible locations and prune them down. 99% of those are gonna be irrelevant. What AI does is AI helps guide the simulation engine or Cosmo Tech simulation engine towards KPI fulfillment so that you simulate as few as things as possible to achieve those KPIs.
So AI guides the simulation engine to achieve the right effect.
– [Ryan] The enterprise AI side of things as it relates to enterprise and commercial IoT, which is like obviously where we focus and the main stuff being talked about right now from, for most people I talk to is learning about more the consumer facing AI applications.
But when you bring it into this commercial enterprise world, the power is pretty impressive of what it can do. And if you tie that information in, let’s say you have access to data that you’re using to help make decisions. It plugs into that simulation, that AI model is able to be utilized, and it just sounds like it just adds more value on top of everything that we’ve already spoken about of how to make better decisions earlier on in the process.
– [Sam] One other thing that I think really important about this world of AI guided simulation is that if you think about the traditional ways of projecting forward into the future, a lot of times what companies do is, step one, take a whole bunch of data silos that exist inside of a company.
Try to normalize all that data, try to rationalize it, and then do simple heuristics where you’re projecting forward based on if this was happening, project it forward. What’s interesting is that you can almost think of that as past data or historical data. Historical data’s pretty messy and it’s tough to normalize and even in some cases, find.
What’s interesting about things like what Cosmo Tech does with the AI guided simulation is you’re effectively dealing with future data. You map what’s right now, and then the AI guided simulation engine goes in projects forward, and so there’s no data to normalize, right? It’s simply the AI guided simulation engine will project forward based on the current state. It doesn’t have to take historical data into account. Makes it much easier.
– [Ryan] Is there ever a scenario where you have, or simulation and the digital twin that we’re talking about here come in after deployment is out in the world as far as let’s say a company has a number of sensors installed in a number of different properties, locations, and then looking to expand from there.
How does simulation play into something that maybe is more established?
– [Sam] That’s a great question. So simulation doesn’t have to be used in the beginning of a project. It’s super helpful to tell you whether it’s gonna be successful and how much, but, let’s take an example of a smart factory and I’ve got a bunch of production equipment. I have a bunch of sensors in this case, it’s an industrial IoT scenario.
And, I started off with operational awareness. I figured out what my overall efficiency is. I asked those beginning questions about why isn’t efficiency better? And I find some of the bottlenecks and I do some of the rudimentary analysis and I increase production by three, 4% or something like that.
Simulation is absolutely useful in this case where you can come in and then map all of the equipment and how it all relates. The people that are interacting with it, the facilities, the supply chain and things like that. And then tell it what your goal is. It’ll find those next, five, 10% of efficiencies of an existing solution.
So yeah, it’s not, I think it’s a great point, Ryan. It’s not just for a ground floor project, it’s also for an established one.
– [Ryan] Fantastic. Last question I have before I let you go here is just what does the future of this space look like? We’ve talked about digital twins here, we’ve talked about AI and digital twins, talked about simulation. Where does this evolve from today as we look out into the future? And what should people be looking for and to be excited about just in the kinda simulation digital twin kind of environment and IoT?
– [Sam] One of the things that’s most, I think, promising about this space is, five years ago, or even 10 years ago, this was only the realm of like super advanced experts that understood how to build AI, that understood how to build simulation, that understood how to combine those. Companies like Cosmo Tech are really democratizing this so that anyone of any sort of skill level can come and use their system and take advantage of AI guided simulation.
That’s really the future is to make, to put this stuff within everyone’s reach. I think everyone probably in your show has either heard of or experimented themselves with things like GPT-3.5 or GPT-4. These generative AI things are pretty fascinating and everyone’s got a really good sense of what that’s like.
What’s interesting is that when Cosmo Tech is taking advantage of AI guided simulation, there is a component of generative AI to it, like it is generating these new future possibilities. The thing that’s different about it is it’s deterministic. Whereas, things like GPT-3.5 and GPT-4 are prone to hallucinations.
You ask it something it doesn’t understand, and now it’s hallucinated. You can’t do that if you’re trying to, if you’re working in an industrial environment, you can’t hallucinate an outcome. And so you can think of Cosmo Tech as taking advantage of deterministic, generative AI in this case to guide their simulation engine.
– [Ryan] Fantastic. Very exciting stuff to talk about. Like I said, simulation was a big thing when I first started in the space. And to hear it starting to come back, or at least, I don’t say come back, but it’s becoming more popular and more discussed now. I think digital twins is enabling a lot of that discussion.
So this is awesome. For our audience out there who wants to learn more about these topics, follow up, engage after this conversation, what’s the best way to do that?
– [Sam] Cosmo Tech’s website is cosmotech.com. I think it’s great way to get engaged in the space and learn more.
– [Ryan] Fantastic. Well, Sam, thank you so much for taking the time. Really appreciate it and excited to get this out to our audience.
– [Sam] Thanks a lot, Ryan. Take care everyone.