dorsaVi Ltd ($DVL)

Hi, everybody, maybe we'll get into it, and there will be a few more that will join as we go. But what I thought I'd do today is obviously, we made an announcement on Monday that I was pretty pumped for. I think it's a pretty significant milestone for the business and mainly from why I came and joined dorsaVi in the first place. But what I thought I'd do is just step back just a little bit, just touch on a couple of slides from a presentation I gave at a conference in Sydney in March. A lot of it is still relevant. And you'll see that there is a plan that we're ticking along with towards the bottom. And essentially, then break from that and show how that fits with the announcement we made on Monday and then field any questions as regards. So if you have some questions, simply put them into the chat. Sam will help me out, and will feed them through to me. I'm happy to answer them as I go. Definitely happy to take questions at the end of it as well. And yes, hopefully, we're all richer for the knowledge of what we're trying to do here at dorsaVi. I hope you can all see the screen, Intelligence at the Ultra Edge. So that's where we're trying to go as an organization. And I'll step you through just why we came up with that, okay? So essentially, dorsaVi, the traditional business, if you like, sensor business has an FDA-cleared medical sensor. For those that don't know, it's about the same size as a traditional domino tile, but it's got quite a lot of technology on it. And they've got several different products that go with that, both with the sensor itself, reading and measuring movement data right at the sensor, but it's also had video information that it marries up with that movement as well. So you can see the sensor captures the data passes along mostly to an edge device like an iPad or similar for some of the more intensive video stuff we do that passes off to the cloud. We've got a bunch of AI algorithms, and we pass that back through and essentially the clinician or the person using the device make some decisions around how that human movement is working. So that's fantastic, but there was a bit of an aha moment, which essentially was that capturing and measuring the data at the source, at the Ultra Edge and passing that through to an edge device or the cloud device probably wasn't enough, and we needed to start moving as an organization to making local inference decisions on the sensor itself. So where that data is captured is where the real power will be. And it was that sort of aha moment that if you like that made them invest in -- made us as DVL invest in, 2 key things in particular. One was neuromorphic computing out of Technion, so Technion is the University in Israel. They've got a fantastic track record for this sort of thing. Neuromorphic is like computer memory, but it's not the same as computer memory. It really wants to mimic our human brain works in silicon. And that silicon really needs to be something that has an analog native edge to it, which is where resistive ramp comes into it. And we have 2 partners that we're working through with the resistive RAM, NTU. So that's Nanyang Technical (sic) [ Technological ] University up in Singapore. They are specialists in the materials, and they have a huge laboratory up there. They can make their own laboratory RM, if you like, and those sort of things. But that's great in the lab, but you need to be able to take that from the lab and push it through commercial foundries globally. And that's where ITRI come in. They are an Industrial Research Institute. They help onboard people into the Taiwanese semiconductor ecosystem. They do that a lot. They [ frontend ] in essentially people in the foundry business in Taiwan. And we all know some of those are the biggest in the world, people like TSMC, GlobalFoundries and the rest. So very, very happy that we've got those 3 partners on board. And I essentially came to DVL to help bring those 2 things to commercial fruition. It's a lovely bonus that we have such a good sensor business that we can take advantage of these technologies. But primarily, when I came across, it was to make sure I brought those 2 things together and brought them into commercial reality, which is why Monday's announcement was important. But I will touch on Monday's announcement in a moment, okay? So just quickly talking about our partners. I think this is really important for investors to understand. We have some of the best in the world partnering with us. So you can see you got ITRI, I've just explained. They're the onboarding partner into the foundry business, Nanyang, NTU. The people up there are just fantastic. So I've been working with them now. I'm on their technical calls. The professor up there, Dr. Lew, he's amazing. He's also part of our Advisory Board. He and his team -- there's a huge team of people up there, all Ph.Ds, all working very hard with this, and they work very closely with ITRI to bring our RM to fruition. And of course, Technion. I mentioned that in Israel. We have that global exclusive license now to Technion neuromorphic IP, and we're actually starting to look at different commercial partners with that and also how we will commercialize that, but equally, how we will implement that into our own sensor business today and how it will look in the future when we can put neuromorphic native into the RM chips that we're developing. Just a little quick overview of them. I won't sort of dwell on some of these slides. I will keep going. Essentially, here is our Advisory Board. So it's really important for investors to understand. So Professor Shahar there in the center of the screen. He's essentially the inventor. I think he is the dean up there now at Technion University. And he and his team, they're the ones that came up with the neuromorphic over many, many years and that we've now got the global license for. And of course, he wants to see it come through to fruition as well. So he has kindly joined our Advisory Board. You'll see there on the right, we have Professor Wen Siang Lew, who is leading the charge up in NTU. They do a lot more. I would like to do more and more with NTU over the coming period of time. But for the moment, we're really concentrating on bringing our RRAM chip through 180 nanometers, all pivoting around the performance of the laboratory 40-nanometer chip, and we'll end up falling at around about 22 nanometers with that RRAM chip, all sitting on top of a CMOS platform, which we'll explain a little bit in a second. So those 2, particularly on our advisory board are fantastic and very heavily involved with them at the moment, delivering on some of the products. But you'll see there's a third person there, Roger Peniche. So he is a robotics specialist and also heavily into advanced manufacturing and that sort of thing. And that's certainly an area where we feel the technologies that we have now through neuromorphic and RRAM and as they come online, where we think we'll be able to play a bit of a role. It's an adjacent industry and market that we want to get into. And you can see there that Roger has fantastic experience in that as well. So we haven't announced too much in that space. I think we did a little bit around exoskeletons, which also fits in there as well. But certainly, there is a -- I've got a fair bit of focus in that space as well as we start to unlock the power of the technologies we have. All right. So just there's some macro tailwinds. I'm not going to dive too deep. We all know this. We all know there's an AI boom, the data center boom. We all know that they need a huge amount of power and water and energy and data centers are building out. We know there's a shortage in the memory chips. The thing I'll say on memory chips at this point is there's a collective of foundry capacity in the world to create both CPU -- so memory process -- memory chips, so processes and memory. They're similar processes. They're made in the same factories, but that capacity is largely capped. So as we develop our 22-nanometer RRAM chip, we know -- well, I know that we will start to -- because it's so good, we'll start to push out others that are currently in the production in those -- that foundry capacity at the moment. So we will replace somebody along the way in the next 18 months or so because we believe our chip is so good. And the reason we go 22 nanometers is you get that really -- it's a good cross-section of manufacturability from the foundries, performance from the chip itself and then the density that you can put on each stick of memory is at the right sort of economic level where we become really useful to robotics, drones, anywhere -- vision systems, anywhere where there's sort of low power, high speed or low latency and low power demand. So we know that, that will come into its own into the future. And -- but it's just important to know that there is more or less a fixed capacity in the foundry supply chain, if you like, at the moment, and the best technologies will get that foundry capacity, and that's where we think we'll fit moving forward. And the other one that sort of increases the business case for us really is the geopolitical uncertainty in the world as well, where you want more and more localized decision-making. And certainly, our memory and computer memory utilizing neuromorphic really lends itself to localized decision-making. And any decision you can push to the Ultra Edge is a decision that doesn't need to be made in a data center somewhere. So -- and it's cumulative, right? So the more and more you can do at the Ultra Edge, the more and more you free up that capacity to do something else in the AI data center. So I think -- and we believe that there will be more and more focus on moving AI out of the data center and into these edge compute devices. So when I say intelligence at the Ultra Edge, I just want to reinforce it's around at the sensor itself. So at the point of gathering the data, have you got the ability to sense the movement or sense the thing that happens, decide, which is transform that data into something meaningful and then act on it. So for the current traditional business that we have in -- at dorsaVi on our sensor, that would be things like you're sensing the human movement, you might be moving to the point of potentially rupturing your ACL. Have you got the capacity to decide that, that's what's happening and then, in fact, act. So trigger some sort of action to potentially protect, said ACL. So they're the types of things that we're looking at on our current sensor. Another one might be you put the sensor on someone's back of their neck or spine or somewhere, and it would be able to sense whether or not you're getting shaken around by machinery underground where there's no real Internet connection and we don't want our employees receiving a concussion or rattling their brain too much. So how do you decide that? It might be that we have a sensor on the back of their neck that will alert them through a buzz or a beep or a sound or send a message to an Apple Watch or something like that or without Internet connectivity deep underground. So they're the kinds of things that we're looking to do with our current sensor that are sitting in this intelligence at the Ultra Edge space, right? So -- and this is just an example of sort of neuromorphic, how it all comes into memory. But what I do want to -- I just want to dwell on this picture for a second, the image at the bottom left under RRAM. So that is a picture from about 3 months ago. That was our test wafer. So it's a CMOS wafer. And on top of that, we -- at the laboratory in NTU, they build our resistive RAM cells on top of that. So that's the precursor for what we're about to do and why the announcement on Monday is so important is we've done that testing in the laboratory. We've compared that to how we perform against our 40-nanometer laboratory test and material. And now we're at the point where the architecture is done, the testing in the laboratory is done. We're confident with the materials that we have. In the announcement, we mentioned that we have 3 candidate materials. They are different from the materials that we use in the laboratory, but the performance is the same on at least 1 candidate and in the same ballpark with the 2 other candidates. And those performance things, the read, write, the durability, how it performs under different heat and cold and that sort of stuff, how many times you can read and write to the said processes and how they perform power-wise and latency-wise. So we've done that initial testing in the architecture, and we're at the point now where we're able to, through ITRI, get some time on one of the foundries in Taiwan and do essentially a tape-out on top of a CMOS wafer like that of our actual 3 candidates at 180 nanometers. So it's a really promising thing. So we've actually derisked the whole process enormously by doing that. And really, we're just scheduling with the foundry to run that first run, and that will be quite the milestone for us, right? I won't touch too much. Just a quick example on this. So why the neuromorphic and computer memory is so important. So for those that don't know, essentially a traditional computing structure, so the von Neumann creates a bottleneck where you have to take from memory, transfer that data through the CPU, the CPU translates that data into something and pushes it back into memory. So that -- and that takes -- when you say here of people mentioning a CPU clock cycle, that's what the CPU is doing. So every clock cycle it's doing something like that. It's taking out of memory into the CPU, passing it through the CPU in a cycle and then passing it back to memory. And that happens just constantly on and on and on and on and on it guys. And so by not going to a CPU to do computational work at the memory, you can save up to 90% of that effort inside a computer, inside memory chip at the moment. So you can see if you -- we start to hook up our RRAM to our sensor, we'll be able to move -- understand where the movement data is in memory, make that translation instead of going to a CPU and potentially push that back to the sensor all without having to ever see the CPU. So an example of that, I think, is on our next slide, on spiking neural net. So in the traditional computing architecture, traditional neural net, we're consistently looking and polling for data to come in and use it. And whether there's anything coming in or not, we're still using memory and CPU and cycles and all that sort of other things. So whilst you can optimize that as much as you like, it's still processing all the signals regardless of what is actually happening in the world, whereas, for instance, in neuromorphic, we'll be doing nothing until an event happens. So a good example for a lot of people is your headphones, which I'm wearing as well, the noise canceling and that sort of thing. So they only do the noise canceling when there is a noise to cancel whereas today, they're doing it all the time. It just on and on and on. So for instance, the neuromorphic right now today would enormously improve the battery life of wearables such as AirPods. And certainly, we have had some outside parties come to us and try to understand how that might fit into their systems going forward, which is pretty positive already from a neuromorphic point of view. Obviously, it gets more important as we combine the neuromorphic with the RRAM chip itself. That's what we really need. I just have a few examples here. And just for all those that know, this particular presentation is on the ASX, so you can pick this up whenever you like and have a look through it. But we gave a few of the different examples. There's many, many examples, which is why I think I've got here -- this is just giving you a bit of an example of whilst originally dorsaVi chased these 2 technologies down because they wanted to put them on their own sensor pack, they then came and got me on board to make sure we bought those to neuromorphic and RRAM together. And it's from there that you go, wow, there's actually -- yes, we will use it in our human-centric wearables world. So this is where we fit in a human-centered world at the moment. But the adjacent markets for this technology once it's there is quite large. Now of course, we're not going to be taking over the entire, for instance, autonomous systems. I'm a big fan of this for our technology moving forward, particularly underwater drones. I think where you have an underwater drone that you need to keep still inside currents. I think having a neuromorphic RRAM chip handling that say still motion is actually ridiculously advantageous for everybody. So -- but whilst it's not the whole $217 billion market, obviously, that's what that industry is doing, we would be a part of that. So we want to make sure that our device is ready to go for all these adjacent markets. So hence, the reason you see us starting to do some work with exoskeletons, that's a natural extension of our human movement data that we collect. Robotics is another adjacent market, whether it's the human robots that we see run these Chinese marathons or not, but it's more the robots that you might see inside a dark warehouse, an industrial warehouse. They're the kinds of things where we can put our sensors on those devices today and help with the sort of human machine separation, that kind of stuff, safety features early on. But because we have such rich human movement data, we will very much be able to apply that to the robotics world, prosthetics world, exoskeleton world. I also think we'll be leaning into drone technology pretty quickly as well, certainly from some of the conversations I'm having, our technology leans. It's essentially anywhere where you need low power, high speed or conversely low latency and local inference, so local decision-making. So these are the things where our technology in the neuromorphic and the RRAM will naturally fit. You can have smart devices that sit out in the middle of a field. These are IoT devices now. They are on but not on because of the ultra-low power and the spiking neural net and the event-driven nature of them, you could essentially have something sitting in the ground somewhere waiting for an event to happen and then spike to life and you're instantly on. RRAM is nonvolatile memory. Our competitors use it as essentially a NAND flash replacement, which our technology, RRAM will also do. But it's the analog nature of how the RRAM works, which will allow us to mimic how the human brain works in a much tighter way and essentially use the RRAM as the basis of an artificial neural net in the future. Key competitive advantage. So we have a strong IP moat. We have strong partners. We have a technology already in our own sensor pack that we can adopt our new cutting-edge technology into. And certainly, that's what we're doing at the moment. I think I'll just go to here. So essentially, we have 3 key strategies underway. You would have heard a few announcements around the first one, evolve the existing dorsaVi platform. So it's a relatively powerful sensor. It's got an FDA TGA approval. It's used in clinical settings, in sports settings and in work healthplace settings as well. We haven't been doing any intelligence on the sensor itself until now. We've been essentially collecting data at the sensor and passing that off to an edge device to do some stuff. But I can tell you now that in some of the announcements we've done, we've actually way down the path of starting to put some of the intelligence onto the sensor itself and making that local inference real and usable, and we'll be unlocking that for our current clients, particularly people like Select Medical over in the U.S. We'll make that available to them straight away, but we are working with a few other partners in this space to understand the power that, that gives them for the technologies that they want to deploy and will allow that. The second pillar there is to develop the 22-nanometer RRAM with NTU and ITRI. So that is what we're doing right now, and that's what the announcement was on Monday. So just to quickly touch on the announcement on Monday. So why I'm so bullish on that is that really is the moment where we step out of the lab. We've done all that lab testing. We've compared it to the 40-nanometer [ QR ] research chip that we have at NTU. And often, those research tips cannot be manufactured at scale, but the performance is as close you'll get to perfect for that type of technology. So we have those 3 candidate materials going forward. We know that [indiscernible] wafer -- CMOS wafer. [indiscernible] architecture complete. Our testing is complete. We know these things will work and now going through ITRI. I can't tell you who [indiscernible] at the back end at the moment. Hopefully, we'll tell you that after we've done the run. But essentially, we'll now do a tape-out of those RRAM chips. We know that they perform well to the lab chip. They are -- it is at 180 nanometers at the moment. That is deliberate. That proves the computer memory, the architecture, the performance, the manufacturability through the foundries. And then from there, we will start to scale down through 40 nanometers and to 22 nanometers. And of course, we haven't just gone with these materials and gone, let's hope that they'll get to 22 nanometers. We pick these materials in particular because we've done that initial testing in the lab to know that we can take it down to 22 nanometers. So it's enormously derisked. So that's why I'm excited. So we've enormously derisked this whole project to this point. Once we've done this run out, we've done our testing, look at -- dorsaVi will probably take some of those computer chips and put them on our sensors and start to play around with what the power of the RRAM can do on its own. And then we'll start to look past that to how we can start to natively put our neuromorphic IP on top of that RRAM chip, too. And hopefully, we'll have a demonstratable sensor that's running these things in the not-too-distant future. Yes. So once we've done that, we'll start to play through the 40 nanometers to 22 nanometers. We're going to see that entire project through. It's really important for us. The other thing that's important in that announcement is it's built on a CMOS wafer. Why that is important is that is the current production process that these foundries work, and it also makes the integration into traditional RAM or a traditional computer much simpler because the CMOS is doing that sort of interfacing for us. But that's not where the story ends for us with this RAM. Our RAM could do a lot more. And it might be at a point which I would like to spin off with NTU is, how do we create a new production method with our RAM so that we don't need a CMOS wafer. And in fact, we can use the RAM to be essentially a 3D cube to create a physical ANN. But that's stuff for the future, and we'll work with NTU to do that. But that's certainly the promise that RRAM can deliver to industry, which is not lost on me. I think that's really important. And then finally, here, you'll see we've got #3, we've got the unlocked neuromorphic opportunity. So we have that exclusive global license now from Technion for all of the neuromorphic IP. We know that we can put it into our RRAM. We will start to do that, but that doesn't stop us now looking for commercial opportunities. We will start with ourselves by putting some, but not all of the neuromorphic capability onto our sensor. It won't be native in memory until we have an RRAM chip running, but it can still do about 50% of what we purchased will be able to be utilized into that memory into our sensor, and that we'll look to do. But at the same time, we have other -- we have third-party from around the world coming to us wanting to understand how our neuromorphic IP may be able to be used in their processes. So we will absolutely be looking at commercializing neuromorphic where we can, but we'll be a little bit selective about the partners we choose as well. But that is an ongoing thing, and it's live at the moment. Final slide really for me, and then we'll start to take some questions. Essentially, this is sort of the execution plan. I hope you can see that we're actually -- in some of the announcements, you'll see that we're actually doing that. So on the Sensor 6.5, so we are doing that right now. We've already started our proof of concepts, and we've already got our firmware updated and we're playing around in our own lab for how we can do intelligence at the Ultra Edge, so making those sort of local decisions on our sensor itself, and that's going really well. So at some point, we will release that as the de facto standard for our sensor pack moving forward. We'll be able to upgrade all our current clients to that version as well. The current sensor can do that. At the same time as that, we'll be looking at how do we upgrade our edge compute software. So how do we make that more user-friendly for people and also how does it take advantage of the new technologies that we'll be putting on our sensor. But that also kickstarts the process of creating an SDK framework and application that we'll be able to provide other people with. So if we provide our computer chip, neuromorphic computer chip, to a third party, we'll also have to provide them an SDK environment where they can use that to program the thresholds and the neuromorphic capabilities on their own. And of course, that will be a product that we would license as well, but it's also a product that we need to make sure we can modify and control our current device. So that's all underway. The next one is sort of designing our next Ultra Edge chip. So our current sensor, even moving to Sensor 6.5, once we've got line of sight, so post this announcement, right? So once we've done this run, we'll be in the point where we're going to be looking at how do we rethink and evolve our current sensor to do -- to be more modular. So we can put more and more sensors on a person, make them smaller and smaller because they'll be doing -- they'll have RRAM in them, it changes everything for us. So we'll make sure we're working through that quite diligently. And that work will kick off almost immediately once we get our RRAM chips. In fact, we're doing the thinking on that at the moment, right? So -- and that will be important to marry up with the SDK in the previous bundle there. You'll see there the next one is the RRAM with NTU and ITRI. So I've been through that a fair bit now. So we're doing all of that now. That's exactly as we said. So remember, I put this presentation out in March, and you'll see there that we're actually delivering against that right now. So all those things on that particular Chevron exactly what we're doing right now. And then finally, the commercializing neuromorphic. So -- it's just sort of my memory there. So this sort of reflex-based intelligence. So that's the kind of power that neuromorphic and RRAM will give you is essentially that muscle memory, the reflex memory, it's that fast. It's faster than a human brain can work. So there's a little bit of a -- is it too fast, but for some of the use cases that we'll do, particularly in robotics and drones and prosthetics and exoskeletons, making fast decisions is not a bad thing and certainly having that reflex and some of the people we talk with around prosthetics, having a human-like reflex on your prosthetic is actually a really valuable psychological thing as much as anything else. So they are the kinds of things that we'll be doing in that space. And hopefully, you'll hear 1 or 2 announcements in the coming times on that. So I understand I've baffled at you for about 30 minutes now, but just good for you all to get a good understanding of where we're going and what we're trying to do and why the announcement was so important on Monday. I haven't seen many questions pop through. So if there are, now it's a good time to ask them. Otherwise, I will try and do this sort of webinar more often, and I'll try to make it more interactive where I get more questions from investors. But I'll just give it a minute and see if there's any questions popping through. But I do hope that was valuable for everybody. Okay. I think there's no real questions. There will be -- you can e-mail me any questions that you may have. I think we put that on the invite to this. I'm actually in Melbourne on Monday moving forward. So happy to -- if people are Melbourne-based to reach out, and I can certainly catch up. But definitely put the e-mails through if you've got questions, I'll endeavor to answer them. And we'll try and do this sort of thing, perhaps around the quarterlies would be a good way to do it. We'll present the quarterly and the information on that. And then we might have 15-minute Q&A with me every single quarterly we do. So that would be my plan moving forward. But thank you very much for everyone's time, and thank you Justin and Sam for organizing this for me.