Podcast #7: Platforms to Applications, Blockchain and AI

Hello everyone, this is Ed Maguire, Insights Partner at Momenta Partners, with another episode of our Edge podcasts. The podcast gives us a chance to dive a bit deeper into ideas and conversations with some of the most interesting thinkers, practitioners, and authors, that we find have great perspectives around connected industry. Today we have two special guests from Davra Networks, we’ve got Paul Glynn who is the CEO, and Joe Quinn who is the CTO. I know that Paul is fighting a bit of a cold, so really appreciate him making the time to come speak with us and share thoughts. 

I’m going to open it up and ask each of you to provide a quick context around your background and talk about what in your experience informs your view of Connected Industry. 

Paul: Hi Ed, thanks for having us on today, much appreciated. Firstly, a little bit of background into Davra; we’re an Irish company, headquartered in Dublin, our focus is very here and North America, we’ve a US sales office in sunny California, the next couple of weeks we’ll be opening and east coast office in Connecticut. We developed I suppose what started as an EP application, Enablement Platform for the Internet of Things, but it’s certainly broadened in its scope quite considerably since our inception back in 2011. So, we’re around quite a long time, we are an IoT platform company, our background is networking, so we’re network guys, we’re used to working in the world of Cisco, Intel, HP, and those guys for a long, long time now. 

So, that very much I suppose flavors how we come to market, we understand the network and the value of the network, and important that the network in any IoT project, and how to use the network technology. We’re also coming from, you could call it I suppose a big data background, in that in our previous company which we sold to Fluke Networks back in 2007, we focus very much on technology like NetFlow, which was big data before there was big data I suppose, because it was all about looking at exactly who was using the network, when they were using, what applications they were using, how they were using it, looking at quality of the network.  

So we’re coming at IoT from an unusual direction where we regard it, and most of our customers and partners regard it as the next evolution of the networking industry, people are just connecting things they’ve never connected before, things like buses, trucks, trains, ATM, vending machines, and they want to connect them so they can collect data from them, and turn that data into some form of useable business intelligence. And our platform basically helps them on each step of that process. 

Joe, what’s your background? 

Joe: Myself and Paul, we worked together back years in the network management industry, and as Paul said, really that’s where we came from and that’s how we made our first forays into IoT, was managing network devices on vehicles, and we realized that the type of traditional network management technologies out there, didn’t really suit for managing these types of devices. A technology we had built a product around was much more suitable for connecting these devices that maybe are remote, or don’t have the same reliable connection back to a head-in. Also, as Paul said, our big data background, we’ve seen these two capabilities as being perfect to look at this new IoT market, and build applications; so, first connect things, and then harvest the data from them, and from there we developed into an IoT platform.  

As I say, probably first of all building out applications in the transport space, and then probably around three years ago, and it kind of evolved as well with the standards and the way those evolved, things like Sigfox, and LoRa, and NarrowBand IoT, where we’ve built those into the platform, and really its exploded the amount of applications and verticals in industries that we can now target. So, as we said, we’re established in transport, but we’ve also been working a lot recently in areas like manufacturing, healthcare, smart agriculture, mining, industries like that, so that’s how we’ve evolved.  

I guess what we look for when we look as to where we’re going to bring the platform next, is verticals that are ripe for disruption, ones that I guess are starting to embrace this digitization of their assets and their operations. We also look at what the ecosystem would look like to target such a vertical, so when we look at what informs our view, we look at the vertical itself, but we also look at system integrated partners that we can partner with to help scale. Ultimately, we always look for ways where we can achieve scale with our platform; so we look at system integrator partners, we look at the vendors that we have, the vendor relationships we have, how suitable those are for a particular industry, and we also look at the customer side, how willing are the customers to embrace this type of new paradigm for their industry.  

If those things checkout for us, then that’s how we go horizontal into further industries, and grown and evolve our platform. 

You touched on an interesting point there Joe, and that’s the extreme need for vertical expertise in IoT. I wanted to go back a bit and talk about how you decided to build a platform; what were the customer pain points that you were looking to solve, and of course, in the early stages of the market, platforms are really the way that you build that foundation for the solutions and applications to come. But could you talk about some of the key customer pain points that you saw, which first prompted you to build the platform, and some thinking about the platform itself, and the decision to build on the competencies you already had, but also to go into this application enablement platform opportunity. 

Joe: It happened naturally. Pretty much every platform starts off as something specific at one point, and they always say every product is a platform waiting to happen. So, we started off with our first IoT solution we called it, it was built on an IoT gateway that one of our vendor partners released, and it was very specific for fleet management, and had done some very specific features around that, that was with Cisco. As we worked more with the Cisco BU and the Cisco sales channel which is enormous, what we realized more and more is that no two projects they were bringing us into were the same. So, we were constantly coming across this case where we had to do customizations, tweaks, adds, changes, even small ones. We realized quickly that this single product type of approach wasn’t going to suit. 

Even taking fleet as an example, you might think fleet management are smart connected transport, it’s very similar, but the differences between ambulances, fire trucks, a solution for school buses, enterprise fleets, public mass transit, they’re large differences from one to the other. So, we realized quickly that we had to do something to help us, to scale across all these types of projects that we were being brought into. So, we were given the opportunity and we had to find a way to make it work. 

What we also realized is that we had a lot of foundations in the product that were generic across all these different projects, like for example the ability to harvest data from the Edge, the ability to store data at scale, the ability to inject business logic into our data. So, we realized that we could make that generic, and form a clear boundary between what’s the platform, versus what we have to customize and tailor from one individual customer to the next. So, we starting forming that kind of platform approach pretty much in the early days, and it was a big decision to take for us because, to make something generic in a platform you pay a tax to do that, but that’s really paying off for us, because we would never have been able to scale by doing single solutions for every single type of customer we were being brought into. So, out of necessity really, we evolved into a platform type of strategy. 

What’s interesting too is, the platform market is quite crowded, there’s been a fair amount of consolidation but there’s also some evolution. How do you think about the role of your platform, and as you see the market evolving, what has changed and how do you see it changing at least from your perspective? 

Paul: I’m going to just add to something that Joe mentioned earlier. Because of our background, because of how we were coming at the market, we took a very different route to market, we’re very child-focused; Joe has mentioned some of our vendor partners, people like Cisco, Intel, Dell, we work with those vendors and we go to market through their partners around the world. These are IT guys traditionally, but they’re now operating in an OT operational world, and that’s a big step change for them, and that probably as much as anything else drove our move towards a platform, because they’re dealing with guys who don’t understand the complexities of the technology, they just want to connect their things so they can collect data from it.  

Whereas our platform very much started as a management layer for very specific used cases, but with a real focus on managing that connectivity of the network piece. We very quickly realized that these operational guys just want to visualize their data. That led to a lot of problems for our partners because, exactly as you’ve said, they didn’t necessarily understand their customer’s business. These system integrators who were selling network equipment, would sell the same network technology to a bank as they would to a retail environment, as they would into a government environment, as they would potentially into a mass transit type environment. So, they didn’t necessarily understand their customer’s business as much as they needed to. 

So, by giving the opportunity for them to work with the operational guys, and inject those business rules, and make decisions about what data was collected, when it was collected, how it was collected, and what audience it needs to be put in front of, meant that our platform became a critical part for pretty much every project. Although from a financial perspective it’s one of the smaller pieces of the puzzle, it becomes the one point that everybody sort of revolves around. I think that’s the value of a platform, and to answer your questions, there’s a lot of people out there saying they are platforms, I think the reality is, they’re not. There are an awful lot of different types of platforms, there are some platforms that are purely focused on the device management, and they do it really, really well, but they don’t understand anything about the business logic, and the Edge compute concepts, and they don’t understand the data analytics piece, and they can’t do visualization very well, but they’re really, really good at device management. 

There are other platforms that just say, ‘Hey, we’re a cloud service. Send us your data and we’ll do our magic on it’; and that’s fine but it doesn’t give the customer any ability to manipulate the data at the Edge, to make decisions at the Edge about what data is important. We’ll take a fleet environment for example, a typical engine produces about 20,000 datapoints a minute, maybe five or ten of those datapoints may be interesting, some of them will only be interesting at certain times, so you may not always want to monitor your oil levels, but if your engine temperature goes too high you need to know. So, being able to work and set those rules at the Edge and say, ‘If this happens, then I want to collect this piece of data’, that’s critical. 

Again, getting back to your question, it’s important to define what is, and what isn’t a platform. There are a lot of point solutions out there that are saying they’re platforms, and there are a lot of very-very good platforms that probably don’t do the application there very-very well. I think we had a couple of years of confusion, we’ve had a couple of years as you say, particularly in the last year a lot of consolidation. We’re now starting to see a lot of companies dropping away, or maybe even going back to their roots, saying, ‘We’re not really a platform, actually we’re just really-really good over here’. I think what we will see in the next coming months, and you know the industry better than most Ed, a couple of the big analysts have reports, and have magic quadrants coming out of the coming months that I think will get some manners on the industry. All of a sudden you won’t just be able to say, ‘Hey, yeah we’re a platform!’ because I think the bigger analysts have taken their time, they’ve watched the space, they’ve seen how its evolved and now over the next three to four months you’re going to see a couple of very interesting market reports that will create an A-team, and a B-team in the world of IoT platforms. 

That’s interesting Paul. I think what we see in any technology market evolution, certainly around software is where platforms essentially become that pivot point, that lever pivot point for value, it really is about enabling the applications. They have to have the platform for the applications to be realized. But I’d like to shift the conversation to some of the applications that you guys are working on, I think what’s so interesting is that the problems that your technology gets used to solve, have a lot of similarities across industries, but also a lot of differences. With IoT I think we’re still in the phase of the market where people are trying to find success stories that will validate, and really help them think through their own problems, in any industry. So, could you share some of the successful use cases, and how you were thinking about the applications that your helping to enable? 

Paul: I’m going to let Joe to talk about some of these cases, but just one point; that’s an interesting comment you made where you talk to IoT industry, there is no IoT industry. In our mind we are very firm believers that IoT is not an industry, it’s not a market you can just draw a circle around it and say, ‘This is IoT’. Similar to e-commerce 15-years ago, or mobile 10-years ago, it’s a new way of using new technologies in existing industries. So, what we regard as IoT in the manufacturing world is very-very different to IoT in the mining world, or the agriculture world, but we’ve just got some great new technology that we can use in all those industries.  

Joe, maybe you’ll talk about one or two different clients where there’s a crossover, it might make some sense. I do think it’s important not to think that IoT is just a thing that can be copied and pasted across lots of industries, you need to understand that every industry, and different people in different industries, and different people in different organizations see IoT in very-very different ways. 

Joe: I would agree with that only from the perspective that, our actual customers in terms of the end-user who purchases a solution, they don’t purchase an IoT solution, they purchase something very specific. The fact that it was delivered through an open IoT platform, the key thing is, they got a solution that matches their requirements, within their budget, and in their timeframe. The key thing as to how we deliver solutions, it’s important to know how we scale, because there are so many different partners, we have different developers; I’ll talk about a few of the solutions we’ve been involved in, but there are partners of ours who are writing applications that we’re not involved in, and maybe we just find out about them. There’re all sorts of great innovation going on in our platform, and its related to your earlier question about how the platform market is emerging, and how its evolving. 

The way we see it is, as Paul mentioned, there’s horizontal platforms and there’s very specific vertical platforms. I think the key thing about it is, the platforms that are going to win are the ones that focus on an ecosystem. It’s not going to be a platform for just purely that platform’s technology versus another platform’s technology, it’s going to be that platform’s ecosystem versus another platform’s ecosystem. So, I think we’re seeing a repeat here of what happened in the mobile industry with Blackberry and Apple; at the beginning it wasn’t Apple’s specific features that bit Blackberry’s specific features, it was 200,000 developers on the application enablement of iPhone and Apple that bit 7,000 Blackberry developers. So, the key thing for us is our ecosystem. 

Now, there’s all sorts of innovation going on in our platform, and as I’ve said there’s applications that are great, things that we never even thought of, that our platform could be used for. Some of the things where I’m particularly proud of what our platform has done, it’s a recurring theme that we see, where there’s an IT department who kind of want to take control, and want to have a strategy to deliver applications, fast, lower the risk, with extreme levels of customization, with fully integrated system into all their other IT systems and enterprise applications, at an affordable cost. I guess our main customer very often is the CIO or the Head of IT in a large organization, who has this problem where they’re dealing with lots of siloed applications that don’t talk to each other, and every time they need a new application they need to engage with the vendor, it’s a big conversation, and obviously a big price. 

One of the best examples we have is a customer in San Diego where they were in that situation, every project was a huge project and took a long time, and may or may not work because they were building it from the foundations all the way up, integrating into all the internal systems in San Diego, and they had to go through that every time. So, we started off in San Diego with one solution, it was probably not the most exciting one, but technically it solved the problem for them which was, estimated time of arrivals. So, using a new technology we could get them to within 10 seconds accuracy to tell them when each train was arriving at a stop, and were using a new technology gateways on the trams, using an algorithm that we ran the GPS co-ordinance through.  

Once we were able to successfully deliver that, then we moved onto the next solution that they were looking for, which was a public announcement system, the ability to, if there was a problem on the lines, they could give audio over the speakers to say there was an issue, you need to get a bus. Because it was also built in a platform we also integrated in with the previous application, seamlessly, that if the train’s rolling up, automatically sends an announcement to let everybody know which train it is, and where it’s going.  

When that was done we moved on and did the telemetry project, we’re integrated with the trains onto the onboard computer, there’s a Siemens onboard computer and we’re able to allow the maintenance guys to remotely troubleshoot a tram. This was a key thing for them because, previously if a tram gets stuck on the line, they’ve got about two minutes to decide what to do with that tram, its more serious than a truck breaking down on the roadside because there’s another tram coming five minutes behind. So, they had two minutes to decide, ‘Can we get this thing going again, or do we have to get it towed?’ Previously the drivers would just ring back to the maintenance guys and say, ‘I’m stopped here, I don’t really know what’s wrong’, and they’d be guessing; ‘Okay, turn it on and off again, try this, try that’. But now the maintenance guys can pull it up on a screen, run a couple of commands and see exactly what’s going on with that tram, they can interrogate the onboard computer from back in the maintenance area, and decide quickly, ‘Yes, we need to tow you’, or, ‘You need to do exactly this to get going again’. And then once we’re able to access that data, we’re now able to gather information on fault codes on the trams, and we can do trimmed analysis to see what kind of issue they’re having most, we can also hook into their SAP PRP system to create work orders for parts that need to be replaced. 

Then we’re onto a fourth solution as well, different things around streamlining operations, but basically the CIO there is executing her strategy, her technical road map on our platform, and she’s just able to move fast and get things done in a much more cost-effective way.  

Paul: I think it’s important to qualify there that there were existing applications to manage the ETA of trains, there’re existing applications to do the telemetry on trains, there’re existing applications to do public announcements, but they were traditionally sourced by operational people in the organization, but they were owned and managed by the IT department. So, the IT department had no control over these applications, no ownership of the application as such, but they had to support it, and they had to host it, and they had to ensure that it was available, and it was up. So, this creates a nightmare for an IT team, whereas by standardizing on a single platform and saying, ‘Okay, we’re slowly but surely going to take over each of these applications, and then we will own them, and then we can be responsible to the city and we can sign up to KPIs, and quality assurance levels, because now we own them from start to finish.  

There was always that, ‘Well, they’re responsible for this piece, and you’re responsible for that piece’, a handover is never clean. But when a city like that can standardized on a single platform, and slowly roll in all their applications into it, then they have control over it, and then they can start to share data. So, there are situations where the telemetry guys would never speak to the ETA guys, but that’s a very large thing; if there’s a problem on the train, that may mean that the train is delayed. That’s critical that those two groups know that, so standardizing our stamp on the single platform and delivering everything from there, allows you to let everyone visualize every piece of data you’re collecting, and you can then start to be creative and smart in how you use that data. 

It’s an irrelevant from a commercial perspective that’s probably less important from a technical perspective, but it makes for a very efficient operational viewpoint I suppose, when you’re looking at everything in one place, or at least everything can be accessed through one place. 

Paul, I think you made an interesting point about the role of the platform, because what’s difficult in industrial IoT, or OT plus IT solutions is, there’s a need to provide cross-communications across typically, very different groups of stakeholders that don’t speak to each other. You provide an example in a city and municipal government, but that certainly applies to larger organizations where you have teams that are focused on very specific areas. But this idea that a platform isn’t just necessarily technology, but it also provides that foundational layer which can cross across, provide that how shall we say; the cross collaboration or communications across these different stakeholders, which has been a big challenge in implementing this first generation of connected solutions in cities, and industries. 

Final thoughts on that from you. How did the lessons that you’ve learned from working with cities which are super-complicated, how do those translate to other industries that are also exploring connected opportunities? 

Paul: I think a city is just a massive organization, and arguably most cities are more complicated than most organizations. One of the big issues with cities is there are just too many stakeholders, there could be politically sensitive stakeholders as well. So, one of the things we have been very focused on, certainly from a technical perspective, Joe has been very-very focused on ensuring we’re very open from an API perspective, we can take data from anywhere, and give data to anything. So, if one person in particular is very let’s say aligned with their application and they don’t want to give it up, that’s okay, they can keep their application which shows whatever it is they need it to do, and they can get that application to feed data to us, and we can then take that into a data-lake, and include it in the mix for making decisions for the rest of the city. Or, we can feed data to them and let them make decisions, taking external data from the city. 

So, that’s critical, not every organization has a dictatorial approach where they come in and go, ‘We are doing this, end of conversation. Everybody just do what they’re told’. So, it is really-important to be able to say, ‘No problem. If you’ve got five existing applications let’s work with them, let’s integrate with them, let’s take data from them, let’s feed data to them; but also, let’s add more value on top of that’. 

So, I think politically and from a commercial perspective helps people offset things like budgets, ‘We don’t want to go and replace this, we’ve already spent a million dollars on its last year, and we’ve got three more years on the contract’. So, you could very easily just go, ‘Yeah, we will continue to work with that’, and that’s what makes a good platform. If you’re too proprietary and too closed, and you’re not willing to share data, you’re not willing to allow existing applications integrate, that will make life very-very difficult for you. All organizations need to be open, to say, this is not a greenfield site, there are very few greenfield sites out there, so we will already have to integrate with things that already exist.  

That would be my view on it. John may have a more technical answer to why that becomes relevant from a customer’s perspective. 

Joe: That example I talked about, San Diego, is unusual in that case, they are a private organization, as opposed to a city municipal. So, maybe some of the differences may not be the best example to compare, but cities typically may not be as strong as an enterprise.  Where I see most of the movement with working with cities is, when you have maybe some regulatory compliance that they have to comply with, and where they’re forced into doing something. Or, if you can tag it onto either an enterprise or an academic initiative, where there’s funding coming through an academic body, or where an enterprise is motivated and incentivized to do some work in the city, because it will benefit that enterprise customer. So, if you can find those types of scenarios then I think you can work effectively with a city. Working directly with a city on their own, as I said, sometimes it can be difficult to work, as Paul mentioned, through all the different stakeholders, and to find an ROI that justifies it and the budget. In that case, if you’re just working with a city on their own, maybe regulatory compliance is probably the only thing that would build a bit of a momentum in projects. 

Paul: Joe, that’s something that I hadn’t thought of. We’re working with a city in the UK, and obviously they have Brexit, and there’s a lot of concerns and issues around the loss of manufacturing jobs. One of the cities we’re working with are doing the exact opposite to what Joe mentioned there; they’re collecting data about traffic on the main intersections around the city, and they’re feeding it to local manufacturing companies, who then feed it into their ERP systems to deliver another set of data for their just-in-time manufacturing processes. They’re able to say, ‘If a vehicle leaves your depot right now, it will take nine minutes to get to your manufacturing plant. If you need it there in five, you’re going to have to book it four minutes earlier’, so that’s a situation where the city are collecting data, and feeding it to local businesses to help increase that relationship with the business, and ensure the businesses stay aligned with the city, and feel that they’re getting supported by the city; because in that particular case these two automotive manufacturers are massive employers in the region, and the city want to be seen to be using technology to support those organizations. 

That’s an interesting example of the public sector working to enhance the effectiveness of the partnership with private business, that’s a great example. 

I’d like to turn the conversation to some emerging technologies which are near and dear to our hearts here at Momenta, and Paul, you and I have talked about this before. I wanted to ask a bit about your take on the applications of AI and blockchain technologies to IoT, where you think the best applications are, how the current maturity of the technologies may be a constraint, and where you think some of the greatest value will come from applying both AI and blockchain, which are a little separate but they’re often deployed increasingly in proof of concepts together. 

Joe: Two emerging technologies as you say, very interesting to us here as well, and a lot of it is looking under the covers to see beyond the hype and say how can we make this real for our customers today. Blockchain is an interesting technology, obviously there’s a lot of hype around blockchain, what it needs are the use cases, and where most of the hype comes from blockchain is the use case that most people know best, which is a speculation engine for trading cryptocurrency. So, that used case is not going to be as relevant for us, but we’re looking for what other use cases can blockchain bring. 

These public decentralized blockchain technologies do have some inherent scaling issues, as you mentioned, it has all the complications of distributed technology but none of the advantages, because you’re not distributing processing amongst different nodes. Every node on the network has to do everything, so that doesn’t scale very well because really, you’re as fast as your slowest node. So, what is more interesting for us is the private blockchain technologies out there, and 80 percent of the blockchain deployments in the world today across different industries are private permissioned blockchains. That’s a technology area we’re looking at now, where you can have a private blockchain between different actors who form the nodes, and they’re going to be able to basically share what we call a ledger between themselves. 

One use case we’re looking at, it’s for a particular partner of ours, we’re working on a multi-model supply chain, end-to-end, basically from the source, to the shipment, on road, at the ports, on rail, over sea, over air, all the way from A to B, and being able to monitor perishable goods for example across the entire supply chain. One of the key things I guess when you have multiple actors in that supply chain, you’ve got the guys who are at the source of that, who are saying they’ve got x-amount of goods, the quality is ‘this’, you’ve got all the different points along that supply chain, and then you’ve got the receiver. They do need to trust each other, they need to trust that they’ve done what they said they would do with those codes; for example, if it’s a vaccine, that the temperature was a certain amount, the humidity was a certain level, at every point along the way. 

So, we can certainly collect the data, but if you’re receiving those codes, and you’re looking back, and we can present you a nice chart, and show you no threshold was crossed, we want to make the integrity of that data completely unquestionable. And the pertinence of us who are looking to provide this service want to be able to have that as a key aspect of the service, that you cannot question the integrity of the data, because ultimately data can be tampered with if people get in there, hack it, and know what to do, that’s what the fear is. So, obviously security is a key point to that, but also the ability to create immutable data records, data that cannot be changed after the fact.  

We’re basically able to stamp the key pieces of information, at various key points along the supply chain, into the blockchain, and then when you go to run your report, all the actors along the chain have their own copy of that. So, if any one individual tries to tamper with data, it’s going to become apparent to everybody. That’s one key thing we’re looking at in terms of integrity and trust. Paul, I think you mentioned the other day there was another example of looking at recording the odometer reading for trucks so that people aren’t tampering that. 

So, a ledger is one part of blockchain, this shared ledger amongst different actors to create trust in the data. There’s also a concept of smart contracts where you can execute a mini-contract, maybe it’s not just a single transaction, but maybe four or five different things need to be true before the contract will be fulfilled, and those four or five different things could become true at different stages. So, we’re looking at smart contracts for that, and again smart contracts is something that does not scale on the public decentralized blockchain at the moment. There is a lot of effort in trying to make that scale, I think they’re getting close, but they may be a year or two away, but for a private blockchain it is an interesting technology.  We’re looking at doing some innovations where we could do some off-chain processing to help smart contracts scale, which means that you can do some small things on the blockchain, but for some heavy lifting you take it off-chain and you throw the result back on when its finished. So, there’s quite a lot we’re looking at there with blockchain, an interesting technology area.  

That’s great, and how about the application of machine learning and AI technologies to traditional connected solutions? I know it’s certainly a broad brush and a broad area, in many respects AI is foundational technology but, could you talk about where you see the greatest value for machine learning applied, and any potential pitfalls or caveats when maybe you’re a business looking to implement a solution for visibility and optimization, and you want to try this technology, are there any considerations that you would highlight? 

Joe: Absolutely. AI is an interesting area for us. There’s multiple places where you can inject intelligence, artificial intelligence into you end-to-end flow of data, one is the Edge, you’ve got these IoT gateways where you can run machine-learning algorithms right out at the Edge, so you can be looking at data coming in off sensors, you can have a smart algorithm that’s constantly learning, and it can make decisions right out at the Edge straightaway.  

We’re doing stuff in smart agriculture where we’re taking data from sensors, there’s a very interesting used-case where we’ve got palm trees, the palm trees must have certain conditions, the conditions must be right in terms of the soil conditions, the environmental conditions, the amount of photosynthesis taking place, so we measure all of that. Basically, it’s serious, because if the conditions aren’t right the tree flips gender, it goes from female to male, and whilst its male it doesn’t produce any crops. So, if you’ve got a plantation with 30,000 hectares of palm trees, 10 percent of them are male, that’s money, you’re losing money there. So, we’re applying machine learning right out at the Edge, to monitor those conditions, and when we see a threshold being breached and when we say, ‘This tree’s gonna flip from female to male if we don’t act here’. You can augment some serious actions, you can turn on some sprinklers, hook into the operational systems to schedule the guys to come out and put some fertilizer down, to give that tree, or that area usually, some attention to stop those trees from changing gender. So, that’s one good example of machine learning at the Edge. 

As you go up to the cloud side there’s also a lot we’re doing there, and there’s two main areas, one is trying to predict that problems are going to occur, predict falls basically; using a lot of data from potentially lots of different data sources, feeding it into predictive models, to predict that basically a fault is about to occur, that’s one key thing. The final thing is predicting what the response should be, so once you’ve realized you have a problem, and once we’ve detected that, maybe an event has come into us, the system has gone down, a lot of the time especially with some of the larger organizations we work with, a big aspect of this is just getting this information to the right people, quickly. 

We’ve had scenarios where there’s been incidents that have lasted for four hours on a customer site, and each minute was costing them ten grand, and that was an incident that took them 45 minutes to react to the incident, and then once they’d reacted it took them 3 ¼ hours to get all the right people, and finally get the right person to solve that incident. So, what we’ve put a lot of focus on is artificial intelligence in terms of learning how to respond to incidents better. Get everybody involved, get the people, the machines, and the data involved to help us resolve this incident quickly. 

So those are the areas that we’re looking at in terms of AI, but I guess if you want to talk about the reality of how this is delivered; predictive modelling is a services-intensive thing, you can’t just copy and paste a predictive model from one environment to the other, even if they look the same. Another example of a use case we had was, where we predicted the probability that a vehicle would need to be repaired within the next 30-days, but the factors and the weight amongst those factors, and the algorithms behind predicting that a vehicle needs to be repaired in a utilities environment, would be very different than a school bus for example. So, predictive models need to evolve individually, in their own environment. 

I mentioned earlier we always looking at how we can achieve scale, and I guess as an application enablement platform we’ve brought those principles of enabling external developers across to the data-science world as well. So, what we’ve focused on, enabling teams of data scientists to write predictive models, so Davra doesn’t have to write the predictive models; sure, we can, we have a professional services wing that can do that, but the real focus is on building up, and enabling an ecosystem of data sciences to do that. So, we have a function called Predictive Model Life Cycle Management which data scientists can use to basically manage the life cycle of a predictive modelling for their customers, which allows them to scope out what a predictive model needs to be able to do, and what do they want to achieve from this AI. 

They have an integrated development environment that allows them to develop their predictive model, then the key thing is to train that model, score that model, and finally feed that model with real-time data. So, we set up a real-time data pipe to that model, and basically, they can manage their predictive model over the course of its life, see it get to the point where it’s accurate enough to put into production. The typical best practice is they have a champion predictive model, which is the one that is in production, and the one that you’re reacting to its predictions, and you always have challenger models as well; so, Challenger A, Challenger B. Every model comes to maturity and then it decays and declines. We help them gauge when the right time is to promote a challenger in and in place of the champion model. 

That’s really a very clear outline and explanation of what’s involved. Of course, you do need to use the right tools, the right approaches, and keep it iterative, but it is amazingly powerful. 

We’re coming down to the last question, and what I’d like to ask both of you is, what are you excited about in the coming years in Connective Industry, what really sparks your passion. Also, I always like to ask if there’s a book, a resource, or a recommendation that you like to share, and it doesn’t even have to be related to technology or industry. But for either of you, I would love to get your thoughts on those topics. 

Paul: Ultimately at end of the day I headline the sales guys, so my excitement always comes from revenues, growth, and where we’re going as an organization. I think this industry, and I know I said earlier that IoT isn’t an industry, but I think this industry is really starting to mature. We’ve come through a couple of years where people are dipping their toe in, and they’re trying to get their head around what these new technologies mean to them. We’ve seen huge maturation in the last year, we have this slide that we show, we just funnel approach where we show where the IoT industry is, and how this huge bottle-neck of a top where people aren’t getting through the earlier stage technology trials, and the proof of concepts, they need something like what we’re offering to help them do that. But what we’re absolutely saying is, we’re getting more and more projects now moving through that phase, into the real phase where they’re saying, ‘Okay, we know the technology works, now we want to roll it out in an area of our business, or for a small piece of our business’, and we’re starting to see a real upsurge in the number of projects moving from proof of concept to real-life. 

If the next step after that is okay, and we really like where its working on this piece of our business, now we have to add it to more of our business, or, as Joe eluded to, in several cases we have to start adding new function features. So, the customer, the more the value they get from our platform, the more assets, the more things they add, the more licence revenue we get. To me that’s the exciting thing, that’s where 2018 has already started off, we’re up 200 percent on last year so that to me, this industry is maturing and its coming together now and is starting to make a difference for people. So, that’s the first thing. 

I think the bible for IoT at the moment is Maciej Kranz’s book, ‘The Internet of Things’; now, I’ll put my hand up and say, Maciej is a good buddy of ours, he was our sponsor way back in the early days in Cisco. It’s a great book and it’s well worth a read.  

Beyond that, one book I always-always recommend to people, it doesn’t matter what size or organization, it’s very much aimed at start-ups and tells you about the mistakes you can make, but everybody should read it, it’s called ‘The Hard Thing about Hard Things’, by Ben Horowitz. It’s fantastic, I sit, and I read it, and I realize all these mistakes I make every day, well, smart people like Ben have been making them for years, so I can’t be that bad! It’s just a great read, it really is. 

We’re coming up to the end of our time here. Joe, did you have any final comments? 

Joe: Just to your question about what excites us, really there’s a couple of trends that are exciting. Paul mentioned the maturity of the market, that’s one thing.  Vendors, devices, gateways, sensors, they’re getting better, we’re able to build better solutions with them, standards are maturing. But I think the most exciting trend I’ve seen is the capabilities of our partners to adopt and build on our platform, so system integrators who maybe traditionally had skillsets in networking, data center, telephony, the ones taking IoT seriously are now building IoT teams where they have developers in-house, and they’re building that skill-set.  

Also, traditional IT departments that want to adopt strategy around IoT, hiring developers in-house. So, a real trend that I see, is that the capabilities of the ecosystem we are targeting is really ramping up, in the last year I’ve seen a massive difference. So that’s really exciting for us, because as a platform provider those are the users that are going to make up our ecosystem, help us scale. 

Again, thanks to Paul Glynn, CEO, and Joe Quinn, CTO of Davra Networks.  

This is Ed Maguire from Momenta, I am the Insights Partner and we will be sharing the resources from the podcast. I want to thank everybody who had stayed with this conversation until the end, it was a fascinating exploration of a lot of practical insights, as well as a lot of new things which I learned in our conversation. So, I want to thank you both, and thank everyone for listening.