 Hi, and welcome to the Antmicro booth. We're here at the MetaWorld 2020, and I'm Michael Gilda, VP Business Development at Antmicro. I'm going to share around the demos here. There's quite a few of them. This is what we call the video wall, which covers quite a bit of the AI-related technology we work with. There's some FPGA and GPU-based ore tracking for mining purposes. There's some devices we build for customers in the medical industry. There's a multi-camera solution here running. As you can see, this little Antmicro logo is big in spinning middle. There's some other hurricanes that we've built, but most of our actual activity is around software that you can put in those. What do you do with those cameras? So those cameras are just a six-camera solution on our custom camera modules. You can build different kinds of multi-camera things like 360-degree vision kits or tracking solutions to our vision. This is an allied vision camera, allied vision is a partner of ours. We build all the embedded drivers for them for Linux-based devices like the NVIDIA Jetson Xavier or the Antmic 6, Antmic 8. Which part of this is Antmicro? So actually, it's some of those boards behind this module, but also the software running inside of this, right? Again, we're a software company, so I would say that the most important thing to carry out of this is, sure, we do a lot of hardware, but in reality what we do is we put advanced computer vision AI software on top of those boards. What's the name? Antmicro? What's it mean? It is what it is. It originates from us originally being a small company that works with micro-technology. That's all. And here is UltraScale Plus with Xilinx? Yes, correct. That's an UltraScale Plus FPGA system on a module. And some more NVIDIA. So this is our open-source hardware board. It's the Jetson nano-carrier board. That's what we originally called it, except that it's also compatible with the Jetson Xaver and X. This exposes PCI Express, so you can connect different kinds of PCI peripherals to it. This here is a four times Ethernet hub switch that you can connect into the nano, getting this multi-Ethernet solution in a very small form factor. And here you're talking about some RISC 5. You've been doing a bunch of stuff with RISC 5, right? Yeah, RISC 5 is one of our key focuses for the future. There's quite a lot of research and actual customer projects we're doing today. This module we're looking at here is a RISC 5 SOM we built some years ago for one of the RISC 5 summits. And then that module is actually plugged into this badge that we built for, I think, the RISC 5 workshop in 2017 where we gave those away to speakers and press. A lot of people actually use those until today. We also inspired quite a bunch of different kinds of badges across many conferences. So what do you do in the RISC 5 ecosystem? Do you work on the chip or the software for the chip? Who's doing the chip? Is it with sci-fi? So there's a few chips here, right? This one here is actually sci-fives. But the one we built in here is, as you can see, if you zoom it up, this is an ant microchip. We call it the Gem 1 since it's the first in the series. This is a system and package built up from several components. And what we're doing here is doing a machine learning algorithm detecting the brightness of the light approaching the sensor. So it links the LED when we have a lot of light and then if you cover the light, there is no blinking anymore. This includes the RISC 5 inside. Actually we have a second version of that chip with a hard RISC 5 core. We have it opened here so that you can see inside. The package contains several elements interconnected by a flexible, yeah, interconnect basically. And this all runs Zephyr. It has the capability to interface, you know, MIPI cameras via custom MIPI-P. And you can also simulate that in Renoad, which is our open source simulation framework that we use for different kinds of embedded system development, but also for hardware software code development, for testing of AGI systems like the TensorFlow Lite. There's a dedicated pod for Renoad related stuff in here. Hey. I think the most interesting thing here is a board that we just announced with QuickLogic. QuickLogic is an FPGA vendor. The EOS S3 is an FPGA SOC, which we created and delivered an entire open source hardware and software ecosystem for. So this board is open hardware. It's on GitHub. And on the board you can run Zephyr, which we also ported to it and released as open source. And this is also supported in Renoad. There's some open source IP. There's some tutorials. And overall it's an entire ecosystem that we've been building for QuickLogic to enable people to quickly start working with the EOS S3. This is actually an ARM chip. This is a Cortex M4 with a flexible FPGA logic inside. So what's the big use for Renoad? The biggest use for Renoad is actually testing and development of software. If you want to build a complicated system with lots of software in it, especially when protocols are involved, if you're connecting multiple devices together, you can take the software you'd normally run on an embedded device, but instead of running it on hardware, you'd run it in Renoad without modification. And by doing so, you abstract out the hardware problems that you normally face. You work with the software in a pure computer infrastructure. So you can do continuous integration, repeated testing, you have reproducibility, you have traceability. You can look at how the software's interacting with the virtual hardware to learn more about why your software isn't working. With real hardware sometimes, you put the software in, it just doesn't work. With a simulated platform like Renoad, you can actually see what's going on and that's why TensorFlow Lite is using us for prototyping development and CI. What are we looking at here on the screen? Here we have a system of three interconnected nodes. They're connected over UARTs and they're just talking to each other in an automated routine. This is real Renoad running on the screen and it's just all automated. And of course you can imagine much more complicated setups, except of course this little device, this low-par device, it doesn't have any super powerful IOs. But we also, in Renoad, we can also simulate platforms like the Polar Fire SoC, for example, from Microchip, which includes things like PCI Express and CAN and USB, which allows you to build pretty powerful and complicated setups with multi-core Linux, FPGA co-simulation and so on. That's another RISC-5 stuff? Yeah, Polar Fire SoC is a RISC-5 FPGA SoC from Microchip. So this Renoad, is it revolutionary? Is it a new kind of way of doing things or is it similar to a lot of stuff that's been done before? We like to think so, it's revolutionary in the sense that it gives you an ability to innovate in terms of methodology of development of embedded systems. So very often people don't really test their devices extensively. The problem is that the more complicated your software stack gets and the more hardware is involved, you will just effectively stop being able to physically run through the same test routines over and over and over again. With Renoad, you can connect virtual sensors, press buttons, see virtual LED blinking. You can even have, you know, graphical outputs if you wanted. And all of that is in a virtualized environment that is reproducible and traceable. So this lends itself very well to CI. Normally, if you want to test hardware and continuous integration, you would have to put, you know, a board or a bunch of boards into a box and hide them in a server room. And of course the question is if you want to press a button, how do you do that, right? So in a simulated system, you can just effectively simulate that you're pressing the button and see what happens. This actually kind of ties very well into the continuous integration and cloud and over-the-year updates aspect of our work. What you can see here is an open-source Android BSP that we built and released to GitHub. And this BSP, you know, the way we work with those things is, it's a lot of repositories. It's hard to manage. So when we work with our customers, we give them reproducible and traceable pipelines and code accessible to them in the cloud and the cloud system that is put together in a cloud pipeline and then out of this, you get a reproducible BSP build that you put on a board like this and you can trace back whatever the software, wherever it's coming from and how it really works. It's a nice platform for collaborating with the customers and for making sure that what you think you're running is the same as what you're actually running. So Antmicro, how big is the company? Where is it based? Where are people based? So most of us are in Poland. There's over 50 of us. But we act worldwide. Our customers are mostly from the US and EU. There's a bunch of cases in, you know, Australia and Singapore. But yeah, the majority of our work actually happens in the United States and, yeah, in different European countries. Over there, there's more FPGA going on here. Is this, we covered this or not? I don't think we have. This is an FPGA platform. It's an open hardware board again. As you can see, we're very fond of open hardware. This includes CSI IP, which allows us to interface cameras. And at this point, we're showcasing the methodology of development that we use for FPGA. So we work a lot with risk-five-based open-source soft system on chips, and there's one in here. There's also Zephyr, the real-time operating system running on this platform. And it's kind of interfacing the camera into the rest of the system, running a shell. And we're talking about, you know, vendor-neutral development methodology. We work with all of those vendors, and we like them very much, but we think that ultimately, FPGAs are a grossly underestimated platform that could be so much more popular if only the methodology was more software-driven. And that's what we're aiming for. We're building blocks and tools and IPs to enable a much more software-oriented FPGA development methodology than before. So you're working on stuff that's ARM-based. You're working on FPGA stuff. Intel, XCD6. I saw an Intel over there, right? And risk-five. Are you all over the place? Yes, I would say that most of our current workers of ARM, the much of the future generation stuff is risk-five. The X86 is not really a lot of what we do, but of course it is, you know, part of the ecosystem. Connect to it. You connect to it, right? Yes. So what would you say is the status of the risk-five stuff and it seems your company is very enthusiastic about it and doing a bunch of stuff about it, right? There's a saying where people grossly overestimate what can be done in five years and grossly underestimate what will happen in ten years. So I think that's what's happening to risk-five. Five years have passed since we joined that ecosystem because we joined in 2015 and I think there's been a lot of progress, and of course some people anticipated things would happen sooner. In reality, I think all this groundwork, all the stuff that we did, you know, in terms of software support, simulation, operating systems, all the building blocks are there and in the next five years, risk-five is really going to make a fundamental difference to how we perceive AGI systems. And why is that? Well, because it gives an open capability of collaborating between people. Like you can see the risk-five foundation being, I think it encompasses around 300 companies today. So this multitude of voices is giving us an extreme ability to innovate where people just bump in each other and work together to build great things like the gem one I was showing here, this video solution there. All of these things are in some shape or form a collaboration between different parties and different open-source frameworks, different pieces of software and hardware. This is very easy to do in an open ISA that is risk-five. But there's also a lot of open-source possible when happening around the arms, right? Sure, no, absolutely. So arm is still, of course, and it's not going away. What is really happening is that new markets are being built, new capabilities are uncovered. And I think that's really risk-five's role is building those new things. It's not necessarily pushing out existing solutions or, you know, killing existing players. It's rather enabling new innovation in fields where it wasn't possible before. And I see the Zephyr up there. So what do you do with the Zephyr? So throughout some of those demos you might have heard me say Zephyr quite a few times. We use it as a default RTOS for all the things we do. So of course in bigger devices we use Linux, so many of those AI demos will be running Linux. But if we go to a smaller device where Linux is not necessary and actually might be an overkill, that's where Zephyr comes in as we call it, you know, Linux's little brother. And is it fun to work with Zephyr? Of course. Yeah, it's a great system. It's a great community. So it seems that at EndMicro you're very excited about open source. Yes. Only open source. Everything is open source. I mean, of course, a lot of the things we deliver to customers ultimately is some kind of a close product that they sell. So it's not all open source all the time, but all the building blocks we use, all the fundamental technologies that we apply are open source. And we contribute a lot to open source flows that need to happen for work to go forward. Just need to be open source. So, yeah, it's almost all open source. But I wouldn't want to say, you know, we can't work with proprietary technologies, just that they have a place as the finer layer, you know, the sugar on top that you might need to make a business case. But for the majority of the infrastructural things, open source is the best way to innovate. And you say Poland? Where is it in Poland? Postnat. The most talented open source engineers in whole Poland? I mean, that's where we opened the company, right? That's where we based. So I can't comment on whether it's, you know, most talented, but definitely we, you know, enjoyed there and it's given us a lot of great growth and opportunity. And every year, like this embedded world, it's a little bit strange one, because a lot of growth in this embedded world, like in the industry of embedded, a lot of new things happening, exciting things. Yeah, I mean, you know, this year is pretty special, of course. But ultimately, if you look at the landscape, yes, we're going forward. Yes, there's a lot of new innovations. I mean, both for end-microbots, for people in general, there's quite a lot of excitement. So happy to be here. This is our seventh year as exhibitor.