How do machines on the internet know who and what to trust? As tens of billions of internet-connected devices are already online and billions more added each year, this becomes a real question and one with no easy answer.
In this episode, I talk with Jason Shepherd, VP of Ecosystem at edge orchestration company Zededa. Jason shares with us how he helped lead an open-source initiative to form an entire platform to answer these questions of security and trust through a series of fabrics and interconnected ecosystems. We also discuss why open systems, distributed computing will win in the future and how AI at the Edge is changing the way the Internet of Things is being adopted.
If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future AppliedAI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!
Resources and Topics Mentioned in this Episode
Enjoy!
Your host,
Justin Grammens
How do machines on the internet know who and what to trust? As tens of billions of internet-connected devices are already online and billions more added each year, this becomes a real question and one with no easy answer.
In this episode, I talk with Jason Shepherd, VP of Ecosystem at edge orchestration company Zededa. Jason shares with us how he helped lead an open-source initiative to form an entire platform to answer these questions of security and trust through a series of fabrics and interconnected ecosystems. We also discuss why open systems, distributed computing will win in the future and how AI at the Edge is changing the way the Internet of Things is being adopted.
If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future AppliedAI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!
Resources and Topics Mentioned in this Episode
Enjoy!
Your host,
Justin Grammens
Jason Shepherd 0:00
So you layer all these insertion technologies to create a fabric. And then the kicker is the framework that binds them together has algorithms in it that give you a score a confidence score for that data based on how much trust insertion technology you add. So you can literally create data in the in the physical world, or it goes through a trust fabric, and then it scores it and it says, Hey, I'm 80% confident that the state is real. And then as part of that metadata, you say, I'm willing to sell this information. This information is private to me. We have sent it out into the wild, and then people can buy it, share it, whatever based on those terms, and they know how real is it.
AI Announcer 0:37
Welcome to the conversations on Applied AI podcast where Justin Grammenss and the team at emerging technologies North talk with experts in the fields of artificial intelligence and deep learning. In each episode, we cut through the hype and dive into how these technologies are being applied to real world problems today. We hope that you find this episode educational and applicable to your industry and connect with us to learn more about our organization at applied ai.mn. Enjoy.
Justin Grammens 1:08
Welcome everyone to the conversations on applied AI podcast. Today on the show we have Jason shepherd. Jason is an entrepreneurial technology leader with unique combination of technical and creative abilities, including driving strategy, managing r&d teams, business development and engaging in outbound marketing. Jason is currently VP of ecosystem at edge orchestration company, the DITA prior to joining the DITA. Jason was CTO for the Dell Technologies edge and IoT solutions division. Jason is also a governing board chair for lF edge and was recognized as one of the top 100 Industrial IoT influencers in both 2018 and 2019. He holds over 40 granted and pending us patents as well. So congratulations. Thank you, Jason, for being on the show with us today.
Jason Shepherd 1:51
No, thank you. I really appreciate you having me on Justin. And it's funny, like you got the data pronounced that right? Because a lot of people say is the data and when they're asked, Well, how does it pronounce? And I said, Well think of a Run DMC song like miser DITA. Oh, nice. Well, I did a little bit of research before this show here, too. So I was able to get it right. Well, I'm super excited to talk about artificial intelligence and AI at the edge and everything like that, during this episode on the we'll touch on that. But first, I usually like to dig in a little deeper, I give a little bit of background on yourself, kind of maybe have our guests talk a little bit about how they got into where they're at today. And what was your path that that brought you to, you know, being VP of ecosystem. Yeah, I'm a mechanical engineer from school. So I went to school for mechanical engineering, I started designing sheet metal and plastic enclosures at Dell, early career and, and then after a couple years, I went to the startup world. And that's where you start wearing a lot of different hats. And one of my favorite mantras is the best way to get a job is to already be doing it. So you just kind of dive in. And so over the years, I started to kind of purposely get more and more towards the solutioning side of things. Software, you started running r&d teams and hiring software folks in at that point at Dell, when I had gone back, and just kind of started evolving in that way. And then in 2014, we started this effort within a CTO org that I was part of that was like, Hey, what do we want to do with IoT, this buzzword IoT. And you know, how we got part of the reason we got it funded from an infrastructure standpoint, as I just went out, and started building ecosystem partnerships, ISVs, that needed infrastructure. So that turned into a whole business there, then, you know, ultimately, I you know, I've got a project called edge x foundry started with a small team there that just hit 7 million downloads on Linux Foundation. So, you know, it's all about building this network effect. And so just over the years, left Dell as a CTO for edge and IoT, and came in as the data has got to the point that if you replaced Dell, with the data on all my thought leadership material, I read the same, I called up our CEO, like, I should come work for you. And then we're, you know, I'm like, I'm CTO now, but at the same time in the valley, CTO is a little ivory tower. It's all about ecosystem. This is the era of the ecosystem. So that's how I ended up here. Excellent, very cool. When you talked about edge x and the edge, the edge x foundry, you want to define that a little bit. Yeah. So edge x foundry. This was basically how do I drive more interoperability at the edge from an IoT perspective? no secret. There's way too many IoT platforms. You know, when we started in 2004, I mean, you know, the market like in 2014, there was like, the joke was, there's 100 IoT platforms, and then it was 300 platforms, and then 450. And we knew back in 2015, that the market was going to go vertical before horizontal. So the people that were laser focused on one use case, because it's first and foremost about use cases, they'll get traction, the big players will spin for a bit trying to own everything. And then around the time that people feel enough pain to realize, Oh, we need open in reparability with domain knowledge on top and necessarily unique hardware and software, then edge x that we've gotten started with these sort of cloud native principles of the edge because everyone was being really embedded in their thinking as things
evolved, like around the time edge x got mature, people started to kind of dive in. I started with the team at Dell in 2015 had this epiphany driving in California, I'm like, on the Magical Mystery Tour, I call it trying to meet all these startups and who's got the right edge strategy, and then IoT. And we just didn't see any architectures that were built with the modularity to bring together an ecosystem. So I called up my team at Dell. And I said, Hey, you know, the CTO team said, Hey, if I got some funding, what if we did this, and that's how it started. We launched it into Linux Foundation in 2017. Interestingly, so you know, talk about that maturity curve, 2017 to 2019, there were like 300,000 downloads of those containers within Linux Foundation, first two years, 300,000, the next year, 6.7 million. So it's this curve of like, I got it all figured out, I'm gonna own everything to Oh, man, this is really hard. We need to focus on more open interoperability. And so that's, that's Jackson, your huge community that's building it. So you know, certainly not just me, but I helped kind of seed it with the team and EOS, which is the base of the data foundation. It's a sister project within lF edge within Linux Foundation. And so part of the reason I chose to come to the data was, I'm very passionate about open, my next big project was around trust fabrics, it'd be good to talk about those. But you know, I mean, looking forward into the future, the real potential is interconnected ecosystems, to driving new business models, new experiences, and you will never ever get there without open, you know, imagine if the internet was closed. And so very passionate about that. So long story short, that's that's kind of the trajectory they've been on for about five years in terms of building the the strategy on top of itself. Very cool. I know you're talking to a huge open source advocate here, both in hardware and in software. So I love the Arduino platform is beautiful that somebody created this back in, you know, in Italy, 15 years ago, or whatever. And they've sort open sourced it and been able to continue to let people build and do some pretty cool prototypes all the way to just the Apache Foundation, right? I mean, everything that Apache is built on is really, really amazing. So great to hear that you were part of sort of getting this thing set up. So yeah, so you got this edge x foundry going, kicking ass at Dell, doing all that type of stuff. And then was there a piece of that then that moved you into working at Zenaida, you said that it was there was it some sort of underlying same technology, different technologies, so are you were about but very complimentary. So we're about orchestration for distributed computing. So that's like, I know you're working in kind of deeply embedded stuff. The beauty of when you have enough memory on a box, you can abstract things through VMs and containers. That layer of abstraction greatly simplifies your life, you kind of have this below that whatever hardware you want. Above that, whatever apps you want VMs containers, clusters, you name it. Evo s AR foundation that our cloud our subscription based cloud uses, from a Linux Foundation, think of it as Android for the edge, very opinionated stack. It's not just Linux, just like androids, not just Linux, it's got an embedded hypervisor in it. But that community is curating that operating system that's got this embedded hypervisor, and all these security things that go down into the silicon. And that runs down to about 512. Meg's of memory in two cores, that's the base, x86 arm, GPUs, FPGAs don't matter. Once you can pay that tax with memory. And you can kind of bridge it up to the fringes of the data center. Once you pay that tax, your life is greatly simplified, because all you care about as containerizing are put in a VM, and you're done. You don't have to think about hardware anymore. Same thing with the hardware makers don't have to think about scaling up individual skews for every app that they want to ship. That's exactly like the Android ecosystem, you create a base, you start building a ecosystem of hardware and software around it and do it in an open way. And then it scales. Yeah, well, you mentioned Android, they have their Android things, you know, thing that they just Can I guess recently, I actually was a longtime Android developer, I was one of the first people in line and built a lot of apps over my career. Any thoughts on maybe why they did that? It feels like, well, it's a very Google thing to do is to start projects, and then just end them. Yeah, you know, Android is a, it's a great ecosystem. It's a solved problem in the mobile space. But, you know, I just don't think that you can slap that into this different kind of more edge continuum, just it's a different thing. headless devices, a whole different set of considerations. When you don't have a user associated with, with every device, it's different, like, no one's there to be like, hey, my email is acting wiggy, you maybe don't open this, the consumer world, like, you know, if you're working on an enterprise, it can just shut down the network, if there's a breach sorry, you can't get your Facebook or your sports scores or whatever, like or do your work. But in the OT world, things go boom, you've operations world things go boom, if there's a hack and and the other benefit, though, in the operations world or IoT is that you can be super oppressive on devices and no one's gonna, they're not going to complain back, you know, from a security standpoint. But yeah, I just think it's it was trying to apply a tool to a market that is just not suited for. The benefit of Android is the ecosystem behind it. And they they hit it at the right time when you need an alternative to the curation of iOS. I mean, let's face it, there's a reason why I mean
Smoking from someone with air pods and an iPhone. But iOS has like 20% of the market share globally. Android has 80. As you probably know, so as I say, open builds a bigger stage for a better show. Yeah, for sure.
Unknown Speaker 10:12
You mentioned trust fabrics a little bit at the beginning, I guess. And since we're talking about things at the edge, maybe define what that is, and some of the positives of being a part of this.
Jason Shepherd 10:23
Yeah. So after we got to Jack's going the community to kind of took off the next big effort that I helped lead, when I was at Dell, and still, you know, kind of diving back. And that was part of the beauty of the beauty of open source, you can work on it from anywhere. We started this thing around trust fabric. So, you know, and how do you build data confidence, you know, across networks, one of the challenges? Well, first off in IoT, or any any kind of market, it has nothing to do with technology, the top two challenges or use case in people, whether it's, you know, the consumer, the operations person, the IT person, the line of business, things like not my job that scares me, I don't have the skills. These are this is the why IoT has taken so long to take off, balancing the privacy that you need with the value that you get. And so the notion behind trust fabrics is when you have these complex, increasingly interconnected ecosystems getting beyond Internet of Things to more of a true Internet of Things. You have to automate trust, you can't take people out to dinner fast enough, one by one to build trust globally at scale, you know, just you need technology's help. And no, it's not just about blockchain trust fabric as we conceived it, and I worked with senior fellows, a guy named Steve Todd, you really helped a lot on this. And then we brought in other people, you know, at Dell Technologies. But trust Fabric is a system level approach to building trust across networks. The framework binds all this together, and is manifest in this project called varium, that we launched a little over a year ago, and incubating you can find it out varium.org, there's a cool video that I produced with my team at Dell before leaving that's kind of shows the future of all these interconnected ecosystems through the stress fabrics. You start at silicon, you do open API frameworks, like an edge x, or you know anything, you add confidential computing, immutable storage technologies, ai for context and different things. ledger technologies just keep track of stuff, they don't tell you whether it's quality. So you layer all these insertion technologies to create a fabric. And then the kicker is the framework that binds them together has algorithms in it that give you a score a confidence score for that data based on how much trust assertion technology you add. So you can literally create data in the physical world or whatever, it goes through a trust fabric, and then it scores it. And it says, Hey, I'm 80% confident that the state is real. And then as part of that metadata, you say, I'm willing to sell this information, this information is private to me, yeah, send it out into the wild. And then people can buy it, share it, whatever, based on those terms, and they know how real is it. And this is important, once you get into all these deep fakes with AI and all that, from a nuclear plant, I probably care about 100% confidence. But you know, if I'm at home, if it's turning on my lights in my house, you know, I'm good with 70%, or whatever, you know, just example. But trust is going to be one of the next big things I mean, you gotta face it's going but then to really scale into the true potential, you have to automate trust. You know, that's what these are about.
Unknown Speaker 13:20
For sure. The Internet in general, I'd like to sort of tell the story, I mean, think about when I first got started to me, it was just I built a lot of websites. So it's just not a brochure where so it was just basically, people being able to take a look at information, then you kind of had this phase of where you had the Facebook's, and the blogging and all the type of stuff to get people sort of sharing. So now they're producing content. They're not just consuming, but they're also producing it. And the internet of things really is now machines talking to machines. And so you know, if I were to pull up your, your blog, for example, Jason and be like, Oh, I know this guy, I trust him, I'll trust what I read. But machines can't do that. Right? They have to be able to validate the API's that they're calling against, or the other things that they're getting data, pushing data, transactional data, all that stuff needs to be able to be trusted and having something on the blockchain, you're right is sort of, okay, that gives us a record and time that this happened. But you still don't really have a whole lot of validity with regards to these machines. And when you have 50 billion IoT devices, you know, on the internet in the next five years, it's going to be a problem. Yeah, for sure. So
Jason Shepherd 14:19
on one hand, an IoT device or a thing or whatever, if it's pulling data from the physical world, as long as you can trust it, it's fat. As long as it's reliable leave sensing and it's trustworthy. It's fact versus you know, people can make up all kinds of stuff as we haven't seen that at all. Surely, you know, a lot of crazy stuff going on around there. And then of course, the the AI enabled deep fakes and stuff, it's freaky stuff. So that's the benefit of machines, but at the same time, if you can't trust them, then there's a whole other problem, especially given the scale factor. You know, it used to be that fraud would happen through someone that's automating emails, you know, like the prince in Egypt that wants to send me a $10 million. But now if you start getting into AI and machines, I mean, there's just massive, massive scale potential for bad stuff that happens, you have to start building the ability to combat AI with AI and really understand that trust factor. It's just a really important topic. And you know, the vision that's in that video@varium.org is further out. But you're going to start seeing I think, these pockets of, you know, kind of trusted relationships form, and then it starts to expand. There's also in that video, there's a reason why we put a service provider in the middle of it's got like vignettes from like, smart homes, and manufacturing, and energy and stuff. Everything's interconnected. In the end, as long as you get value, you're willing to give up a little privacy five years ago, I'm like, Hey, guys, you know, we're not gonna do IoT for consumer, Amazon's gonna win calling it right now. Why? Because I sell content and more importantly, stuff. And I've built a relationship with that consumer over years, like, I know my ups driver by first name. And so in the consumer world, people's trust revolves around large companies, large entities, or just any entity that you trust, and you'll give up some privacy to get value. But to scale across b2b to C, or x to x to x, you cannot have one company own the trust. And that's why this whole concept is around decentralize it and make it into these fabrics. You have your circle of trust, important. Also, as a trust fabric is not one version, you assemble it from the ingredients. However, it makes sense to you. If you use open source for everything most transparent, it's probably going to drive higher competence scores. Say you replace one component with a commercial thing, maybe the competence goes down until that's trusted. It's exactly like a circle of trust with people is like you get a new person, they look a little you know, you don't know them, you're going to kind of hold back some information until you get to know them. It's mimicking human trust circles,
Unknown Speaker 16:47
which on a conceptual level that everybody understands, that's kind of sitting out now Can these fabrics Can you have multiple sub fabrics?
Jason Shepherd 16:54
Yeah, that's all intent. So say I'm a business, I build technologies. Throughout my networks, you know, that I own I've built this fabric, I've got these different layers, that creates a context around the confidence of data passing through my fabric, say it's a supply chain, there's another supplier that has their set up. And it's it can be different again, and we just have to agree through open collaboration, what these algorithms are, and this whole point of getting it out into the wild and build standards around it. If my fabric does this kind of confidence in mind is this when they intersect, it creates a normalization effect. And then you come up with a intermediate confidence score. And so it's based on these single abstract, but it's based on these complex relations. And the exact same thing happens in the business world. My organization has a certain level of trust among my people. Yours does we get together, we're a little standoffish at first and trying to fit and then when you reach a boundary condition. And so the same thing will happen with this stuff.
Unknown Speaker 17:47
Very cool. Yeah, I view like a note nodes on a network, right. And there's obviously something that's going to bridge that divide between all these nodes. And maybe and that's that's what you're saying, I guess if you have two different organizations, they obviously need to agree on some place to meet in the middle. But that does happen in these trust fabric.
Jason Shepherd 18:03
Yeah, that's the concept to there's some stuff that I'll varium.org should probably touch on some other top. But so downslide brought in Iota, the Iota foundation. One of the you know, I think leading they're doing some really interesting stuff with their tangle protocol. But it could be any ledger. But we started with Iota. And so dal Iota and Intel recently did a podcast or a webinar on it, like as an update for the project and a lot of interest in it. So you can find that online, you know, just cool stuff. And so we're working on incubating it and stay tuned, but definitely a very important conversation for the future. Just the trust one in general. Yeah, I
Unknown Speaker 18:38
publish like liner notes with each one of these episodes. So I'll be sure to include links off to some of these things as well. Now, when we're talking about these networks, and stuff working at the edge, obviously, there's compute going on. And, you know, I typically like to ask people, how would they define, you know, ai in general, but I'd like to ask you, how do you would you define edge AI?
Jason Shepherd 18:58
First off AI in general, and obviously, very, very important emerging space. And you know, things are accelerating seems like by the day, but then there's a lot of people still talking about AI is like, Oh, I do AI and it's really just like a fancy If This Then That rolls engine, like min or no, you don't do. So clearly a lot of cool stuff happening. Edge AI is basically it's typically in printing, you know, so you're training in the cloud, and then you push some sort of inferencing model down closer to the action, we are seeing a trend towards federated learning where you're kind of seeing a blend even training at the edge, but then you could get regional bias. So even the puresense is training in the cloud and and you deploy a inferencing model out in the field. The Edge is a continuum, there isn't a single one. It's from very, very constrained devices all the way up through sort of regional and access data centers. There's actually a pretty good paper within lF edge. The taxonomy paper we put out last year that describes edge in great detail based on inherent technical trade offs and not loaded terms like near or far thin and thick and all It's like, it's a latency critical, you're always going to run on prem. So here's the on prem user edge, is it latency sensitive, but you want to skip lots of people, you're going to run it at the service provider. So it's the land or land, you'll never deploy your airbag over a win. No matter how fast your your 5g is, there's a lot of people saying, oh, you're gonna drive your car from the cloud with 5g and like, You're insane. It's not gonna happen now augmenting with with services around augmented reality, you know, infotainment. Sure. Yeah. Are you in a physically secure data center? Are you not? Very different security implications? Are you so constrained that you must be embedded? You're kind of going into the Arduino and beyond? Or are you able to abstract it, and that's based on memory. And so as the data as an example, our edge is from that lowest limit 512 Meg's and two cores with EBS from lF edge up to the data center, but we're not in any way trying to compete with VMware Nutanix. Those are solved problems. We are seeing Kubernetes come down and we're meeting it. So Ajay is about deploying it across the continuum. Generally inferencing, a lot of the large infrastructure players will say, oh, everything happens in the data center AI is happening on servers in the data center, because it's basically what they sell. As you know, there's a lot of tiny ml is spinning up, you know, machine learning, like on very, very slim devices. 15 fixed function, but it's happening everywhere. I've said for a while fixed is the new mobile, people talk about ambient compute. But you know, you see more compute everywhere embedded many ways. IoT is embedded computing getting scale. So it's interesting stuff. But yeah, it's inferencing. Some training at the edge. I joke. It's like deep learning in the back and shallow learning at the edge. But yeah, mostly inferencing. Today, gotcha.
Unknown Speaker 21:39
What would you describe as some of the biggest challenges, I guess, that are going on with edge AI.
Jason Shepherd 21:44
Versus use case me a lot of people chatting, computer vision is the killer app for edge AI. Unless you sell Wide Area connectivity, you do not think it's a good idea to stream 4k video over the internet, deploying in the real world, say you're talking computer vision, what works in a lab with, you know, cameras and lighting and angles, doesn't always translate into the real world. Just scaling models. I mean, I'm working on various different AI consortium efforts. And everyone's talking about Oh, AI and the models and all this cool stuff. When they get to actually deploying in the real world. It's like, Oh, that's hard. And so that's my, that's what zita does. But so we think that's a big part of it. I think one of the big challenges with edge AI, in general is that when you get out into the field, there's highly diverse use cases. But then also skill sets. More importantly, the people that understand certain industry domains are not the same people that understand data science. And so you have to kind of span the gap, when it comes to AI models for things like object detection, is that a car is that a bicycle is that a gun or a weapon, you know, whatever, that's gonna become commodity over time. And it already kind of is, there's no differentiation long term, there's gonna be a set of models that everyone uses. Over time, the real value for AI or any of these technologies is going to be people that program models for very specific contexts, or get into the deep, the tiny amount, which is not easy with constrained devices, but I know how this factory runs, and I'm going to work with someone, you know, that does data science, and we're going to develop models specifically for this. This was a colleague at Dell that did work on this zero downtime factory initiative for robots. And there was this guy named Brad that was on the floor been there for like years, and he knew just based on gut feel, like, if it looks like that, that's bad. If it looks like that, don't worry about it. And so they'd brought data scientists in, in Brad consulted with the data scientists and and they created these models, and they call that Brad analytics. And that's one of the challenges is understanding specific context, deploying it in the real world, all that kind of stuff is the reality if you're in the data center, and you're centralized, and you have the right skill set, it's not that it's trivial by any means, but it's just a lot easier.
Unknown Speaker 24:07
It's a lot easier than having to manage these 10,000 devices out in the field and having them do the right thing at the right time dealing with physical that's the thing is I mean got boy sensors can be wonky, right and you can have different lighting conditions like you mentioned temperature can affect some of the stuff you know as well so yeah, it's it's a much different world when you have the physical meeting the digital out there.
Jason Shepherd 24:29
Yeah, I mean, ai in the backend say I'm doing things you know, that are analyzing long term trends. You know, it's financial trends, you know, could be medical collaboration. No one's gonna die if it goes down, right that your vehicle dealer using AI for for augmented or autonomous driving, you're gonna have some problems if that fails. Any issue in the operations world in the physical world causes immediate loss to production and potentially life. Any issue in the IT world has mattered Scale over a long period of time. I've said for a while, like IoT starts in OT and scales and it, it has just a different perspective, a different scale factor. But that's another edge. I said, if you if something goes wrong or you get it wrong, it's going to impact production or safety right now.
Unknown Speaker 25:19
Yeah. Well said, Are you seeing that a lot of companies still don't get it? I mean, to me, it feels like this is kind of a duh, of course, you should be doing it. And I feel like as I talked to more and more companies, or as I've talked to companies over the past, you know, 10 years or so, they're just kind of stuck in their ways, right? They don't, they don't think that these things need to be built intelligently. So data, obviously, you guys have invested a lot of money and a lot of time, a lot of effort, sort of building out some really cool technology. But what what's your general sense? Are people still are we still another five years away from this really taking off?
Jason Shepherd 25:50
I think so I mean, all new technologies or solutions looking for problems up front. And there's been great strides, and like I said, things are accelerating. But we have a lot of customers that come to us, you know, they don't want to hear about edge and they got business problems, you know, and then these are tools to solve those problems. And the reality is, I mean, Zinnia, great, great value wouldn't have come otherwise, I think we're doing something really important and the right amount of early because that just becoming a hotter and hotter topic, as more and more data is out there. And you can't send everything Central. A lot of customers, you know, the reality is they come to us. And if they're early in their endeavor, or whatever, like, they don't get what we do. They're like, Oh, you know, I can do all that, like, apps? And what's my app? And it usually it's like, come to me don't understand that you go tinker with apps, and then you get past a PLC, and then you're like, holy crap, how do I scale this, then they come back, and they're really interested in the way we do it. And it's really flexible. But a lot of people are still kind of early. There's also a lot of legacy thinking is, as I think you said, People kind of trapped, the innovators dilemma, if you will, one of the big challenges in the market, you know, the way I summarize it, and people are having this struggle, you know, 400 IoT platforms, but most of them have never set their foot in a factory floor, the developers, oh, I can do AI, I can do all this stuff. But do you understand how machines really fail? No, that again, that's domain knowledge part. But there's a lot of challenges around established markets, the incumbents, people have been around for a while they have the channel, they have the relationships. Meanwhile, the challengers have the new technology, and the incumbents rarely lead the charge into the new way of doing things. The challenges need the channel, the relationships, the incumbency, the new tech, and then but then they're all like, standoffish because like you're threatening me. And you have to bridge this gap to to solve problems. And that's part of the reason why, you know, from a video standpoint, why we started with VMs. First with the open source component, because we can support legacy apps, and infrastructure right next to modern containers like AI. So take an example, in the safety and security market, video surveillance, large channels, people that know how to roll cameras, you know, drive camera, trucks, out install cameras, no problem. If you're in the data center, like massive Stadium, or a school campus or whatever you can afford, like heavy data center stuff, all these cameras going through network switches to that stuff. If you're doing like 1000 McDonald's or something. The precedent today, as I go out, I install a Windows based video management software. It's all Windows based today, I put it on a box, I connect the cameras, I leave and I never see it again. And so we're starting to work as the data with these this channel. Hey, guys, you can do exactly what you did today, plop that Windows based thing on whatever box you want. And then Meanwhile, as you evolve, you can start dropping AI models on there to start doing facial recognition, license plate detection off that street. And it gives that transition path from legacy to modern. And that's important to bridge the challenge a catch 22 that people are facing in the market. Interesting,
Unknown Speaker 28:48
for sure. And that's all with with containerized models, right or containerized solutions.
Jason Shepherd 28:53
Yeah. So well, Ito as you know, this, the Android of the edge that we contributed now leverage a lot of people are using it as supports both VMs and containers, you can drop on a VM with a Windows based app or some Linux image, whatever. And then right next to drop a Docker container of an AI model. And that's really important thing. Most solutions today are starting with their container. Great. But then you forget about all that legacy out there. Yeah, I
Unknown Speaker 29:17
mean, an organization that has 1000s of cameras already or 1000 of these boxes already out in the field. You're right, they need a slow path to it.
Jason Shepherd 29:24
Yeah, an example. I met with a lot of companies over the years and it kind of that thing around around the innovators dilemma. And I was meeting with a large payments processing company. It was around kind of IoT, they wanted to talk about IoT strategy, and it was pretty clear that they just they didn't know what to do. And they're like, oh, man, when square came out for mobile devices that really impacted us we don't want another square to happen. So we're having this conversation and it was just kind of spinning and and I plan this before the end of the call. I said, you know and I was I had to go in like, Hey, guys, have you thought about when machines start making payments? That blew their minds. They're like, what? I didn't say this. But the fact that that blows your mind is why square is going to happen to you, again, you're not getting out in front and kind of redefining the problem, or the opportunity, I guess, it's not a way to say so.
Unknown Speaker 30:14
I joke with some people, I say, you know, like, one day, you're gonna come home, and there's gonna be a park for your, for your furnace sitting there on your front step, you're gonna be like, I didn't know, I didn't know I needed this. And it's gonna be like, Oh, yeah, you did, the furnace knew that you the fan was about to die. And I went ahead and ordered it for you, right. And that's going to happen. And if companies aren't, don't embrace that, or understand that, it's going to be all about uptime service, and machines taking over a lot of these these mundane tasks, you're right, they're gonna get left behind, they're gonna get squared,
Jason Shepherd 30:42
this, this comes back to the trust factor to have this privacy value curve, gonna value your privacy, you'll give up some, you know, I think that we haven't really faced some major privacy breaches where peepees are on behavior. It's not just companies that have to be responsible for other people's privacy's people. But you know, a lot of people don't understand how much data is flowing around, back out in the web about them, but the dark web, whatever, but, but to get to that were okay with, you know, my air conditioner to talk to some systems and trigger all those actions, I'm going to have to trust that the whole supply chain, back to the trust fabric thing, I think, but I totally agree that's going to happen. advanced class. So predictive maintenance, you know, hot topic, a lot of people talk about it, but then few people actually have the knowledge to go do it. But advanced class would be I don't just know when that machine is going to fail. I know that if I'm going to send a tech out to this region, I'm going to make a route for all of the machines that I need to upgrade and no, no, no, don't fix that one. That one's near the end of its service life replace that one? And no, no, that kind of prescriptive analytics. It just keeps layering on top of itself. But But I totally agree. This is where we're headed. As long as you maintain the privacy value curve. Yes, for sure.
Unknown Speaker 31:53
Oh, cool. Man, we've we've dug into a lot of some really interesting, futuristic concepts and technology and stuff like that. I'm curious how you relaxed outside of your professional life. I, as we're doing this interview here, I can see through the virtual nest, it looks like you're a musician, huh?
Jason Shepherd 32:07
Yeah. Now, this is my office by day, studio by night or the other way around, depending on, you know, kind of working conditions. But yeah, I've been playing a long time. And I joke so I live in Austin, here in Texas. And it's a big music town. When you come here, they just gave all this stuff to you. You want a guitar to take another one. here's a here's a drum kit. It's been great like over the just with crazy COVID times and just all the unfortunate stuff. My current band, we've been able to record remotely and cut everyone's setup. We've been releasing epds, and, you know, bonus tracks and all that all last year. So it's been fun.
Unknown Speaker 32:41
That's good. Do you have any particular instrument you'd like to play the most?
Jason Shepherd 32:44
I play guitar and sing. It's all my stuff. I'm gonna do a shameless plug. The band name is Bella diver. And there's a lot of good stuff out there on that. So divers calm but I'm really fine with kind of like along the lines of, you know Wilco or Mumford and Sons Lumineers, you know, some of our kind of replacements and things like that lyrically driven song, you know, kind of Americana type stuff. It's fun. It definitely helped us maintain some sanity last year.
Unknown Speaker 33:09
Sure, give us a little break from all the craziness and all the work. I mean, again, I I actually played drums for many, many years while I was in college, and even after college, and I feel people think of software engineering and sort of like stem based careers as being so rigid. And I'm like, it's all about creativity, man. It's not it's, you know, I think there's a lot of people that do musicians or artists that get into technology.
Jason Shepherd 33:31
Oh, yeah. I always want to be creating something, whether it's like a new kind of market, community driven stuff, building our house, writing songs, like, one of my many mottos is, if it's fuzzy, I'm on it. Cool.
Unknown Speaker 33:44
Do you have any, I guess, books or conferences, you know, I mean, I'll include some stuff in the notes. But I just didn't know if somebody was getting into this field of sort of tiny ml or edge AI, there's interesting places that they should look.
Jason Shepherd 33:57
What's tricky, because I learned by osmosis from a lot of stuff, you know, it's not like kind of one outlet. I think, in general, I mean, obviously, you are passionate about this, as well as get involved in open source. Open source is a modern way to drive standards. There's a lot of cool projects, there's some stuff around AI within nation and elsewhere. I think standards are going to be important here, like onyx is an interesting open source project within Linux Foundation around how to use standardized frameworks. I mean, there's a lot of conferences and whatnot, I honestly tend to find a number of companies, I mean, a lot of value, a lot of networking, but then also a lot of rehashing of the same concepts. So getting involved with people that are passionate about it, you know, it's there's a bunch of interest groups, there's some stuff even on you know, like social media, around AI interest groups, and just exchanging ideas and things like that you build community. I think you get a lot further than trying to target some book or some I mean, not to down on that, but it's all about playing off of each other and building a community. It's I mean, to kind of back to music. It's like how do you get Good music as you rip.
Unknown Speaker 35:02
Yeah, for sure. For sure. Well, yeah, we I started Uh, well, this is conversations on applied AI. So yeah, that's what this podcast was. But we actually have a monthly meetup that we meet the first Thursday of every month, virtually. So yeah, I definitely encourage people just to what's interesting now is, is everything's virtual. So like, I could attend a meet up on AI in Austin, you know? And then, you know, tomorrow night, I'll be in Boston. So it's just there's all these one groups that are going on everywhere. So that's great advice.
Jason Shepherd 35:28
And maybe you're not those people on video will be real. But no, it's been great. Like from that standpoint, I mean, I I worked at home already. Disney does in California. And so it was kind of a great equalizer for me as people started to realize, Oh, this is what it's like to work remotely like get a little lonely because you missed the water cooler talk. I think it'll be great to just see people again, I think it's gonna get there's gonna be hybrid. It's not like it's gonna go back to normal. I was I think we've all been talking about but without going to see if less of let me just travel just for that. Just to meet somebody For this reason, but it's still good to go meet people and of course, real conversations happen over whiskey. So
Unknown Speaker 36:05
yeah, so we definitely will get back to that a little bit of both. Well, cool. Is there anything else Jason that I maybe didn't didn't touch on that you wanted to end off with? I for sure want you to plug yourself in. You know how people should reach out and get a hold of you. If there's anything you know, is is LinkedIn the best place for you on Twitter.
Jason Shepherd 36:21
I should be doing a lot more on Twitter, but I my my handle speaking of music, former bandmate years ago, nicknamed me deaf Shepherd in my last name, shepherd. So I'm at DEF Shepherd, playing off the hairband theme, but on Twitter and then LinkedIn, I probably do a lot more on LinkedIn. And there's a lot of stuff. I've done a number of blogs and panels and podcasts and stuff. So if there's any interest in my crazy ideas, that's why you call yourself CTO has any plausible deniability. If it doesn't happen, oh, that wasn't on the roadmap. That's just but this stuff will happen. I mean, it's all thing around trust. It's, it'll take time. But the net message there is, don't get locked into some siloed. platform today. And maybe easy, like easy button, I just grab on to some platform turnkey, because if you don't start with an open Foundation, yes, you want hello world fast. But you also you can start with an open foundation. That's that's what we're trying to do. For example, from the data. You don't start with an open Foundation, you will never get to the promised land ever. You'll miss out on all that other opportunity. Because if you're trying to, you can drive efficiencies in your business, but there's a floor. If you're trying to drive new business models and experiences the sky's the limit. So don't try to run yourself into the ground. Think about how do you change the paradigm to you know, it's not one or the other? It's it's but just don't think about zero sum game. Sure,
Unknown Speaker 37:40
said everything's better when it's open. So it's kind of like a rising tide lifts, lifts all boats, I guess, through these ecosystems.
Jason Shepherd 37:46
Yeah, one other thing I know we're going to go but another analogy is when you get caught in the Riptide, most people will you know, common thinking gut reaction, I'm going to swim to shore, and then people drown. And so a lot of people think I just need to do what I do better. I don't want to swim to shore. I'm just going to get laser focus. And what you're supposed to do is swim sideways. And so we need to be swimming sideways more.
Unknown Speaker 38:07
great analogy, Jason, for sure. Well, thank you so much, again, for taking the time to be on the program here and wish you the best and for sure. We'll definitely keep in touch. Maybe I'll see you next time we get to Austin.
Jason Shepherd 38:19
Yeah. And we're in real person,
Unknown Speaker 38:20
South by Southwest whenever that opens up again.
Jason Shepherd 38:23
Yeah, that'd be great, man. Well, thank you so much.
Unknown Speaker 38:25
All right. Take care. Thanks.
AI Announcer 38:28
You've listened to another episode of the conversations on applied AI podcast. We hope you are eager to learn more about applying artificial intelligence and deep learning within your organization. You can visit us at applied ai.mn to keep up to date on our events and connect with our amazing community. Please don't hesitate to reach out to Justin at applied ai.mn if you are interested in participating in a future episode. Thank you for listening
Transcribed by https://otter.ai