Conversation with Zach Shelby
Good day, this is Ken Forster, Executive Director of Momenta Partners and Momenta Ventures, with another edition of our Digital Leadership Podcast. Today it’s my distinct pleasure to interview Zach Shelby. Zach is an entrepreneur investor and technologist in the embedded space, with a passion for tiny ML and internet engineering. He’s a Co-founder and CEO of Edge Impulse, enabling developers to create the next generation of intelligent devices with embedded machine learning.
Prior, Zach was Co-founder of Sensinode, exiting the company to Arm in 2013. At Arm he served as Vice President of Marketing, and Director of Technology for Internet of Things. Zach founded the Micro:bit foundation in 2016, to bring the brilliant educational work of the BBC to children and teachers around the world. Used by millions of young people in over 50 countries, Micro:bit is paving the way for a whole new generation of makers, and IoT pros.
He’s known as the pioneer in the use of IP, web technology, and low power networks, with 6LoWPAN and COEP standards development, and is co-author of the book ‘6LoWPAN The Wireless Embedded Internet’. He has served as the Technical Advisory Board, and Board of Directors at the IPSO Foundation, and was awarded the Nokia Foundation Award in 2014 for his work on the Internet of Things.
Zach, it’s a great pleasure to feature you on our program, welcome.
Thank you Ken, it’s so good to be here.
So key topics we’ll talk about today generally in theme of the conversation, one is of course we always love to hear about your digital industry journey and perspective, we want to talk a bit about your experiences as a founder who has successfully exited your company, a lot of Momenta Ventures portfolio companies generally appreciate those kind of topics. Of course, we’ll want to deep dive on Edge Impulse which is doing some really great things in the space. And then a little bit talking about you as an investor as well.
So, with that in mind, let’s go ahead and jump into your own professional journey. Tell me a bit about that journey, and how it has informed your views overall.
Well I was born at a really lucky time in the history of technology. I got hold of a really special computer when I was a child, at 8 year’s old I got the leftover Commodore 64 from my father’s auto repair business. That Commodore 64 for me was magic, because you could create with technology; I got very frustrated as a kid that you couldn’t take things apart and really understand how they worked, and how you could do more with them. So, getting hold of this Commodore 64 was just mind-blowing for me back in the 1980’s.
I was born in a time where personal computers started to be available to the middleclass in the United States, and that got me started on my path towards computers, engineering, invention, from that Commodore 64 experience. I was also lucky because in the 1990’s it was the time when the internet was starting to transition from an academic network, so it started as a university and military network, to just getting opened up to the public. I was at the very early-early edge as a teenager hacking the early internet servers, breaking into university networks to look at boring syllables and FTP servers, there wasn’t really much there, but it was an exciting time to be using a new networking technology that connected all computers. So, I got very excited about the internet and wanted to spend my career working with invention and the internet, so what drove it.
I went to school in Michigan, but very early-on I felt a need to go see the world and go abroad, I think a lot of young people do. So, in my second year at university I went to Finland of all places in the world, and I never came back! I went to Finland and to the horror of my parents I never came back, I fell in love with the technology boom that was happening over there, and ended up staying, finishing my degree and getting my first job in technology straight out of electrical engineering studies. That first job was wiring up a weather station, I always remember this weather station at the top of one of the research buildings there in Finland, and my job was to make this weather station in 1998-1999 internet connected with a little embedded processing card that was on it. We did that, we made one of the first internet connected weather stations back then, which is still running to this day, that same weather station is out there on that roof, and it’s still operating with a couple of hardware upgrades in-between.
It’s interesting, I consider your early experience to be what I think as full stack, and this is one of the things that we always think about in Digital Industry practitioners is, they’re usually coming from an electrical engineering background, or computer science, and they’ve ultimately touched both the Adams and the Bits if you will, to bring them together in terms of full solutions. So, having a little bit of this background in terms of your own journey is fascinating, especially starting off with a Commodore 64, I actually started off with a Trash-80! So, I can appreciate where you came from in terms of the early days.
Your thread of experience has really migrated up the technology stack, from Edge Wireless to cloud, and thinking over the various companies that you’ve worked with. What are some of the trends that you’re seeing relative to each of those areas, and ultimately thinking how these will converge in some of the more recent activities you’ve been doing as well?
Well something that’s always fascinated me is embedded, how we can embed, compute, and sensing, control, and communication capabilities in the physical things. Even though I love computers I don’t want everything to be a computer, in the sense that I have a keyboard and I have to spend my time and my energy fighting with it, because we do fight a lot with computers we have to admit. I love this kind of integrated embedded technology that’s just there, it works, and that’s what fascinated me in my early career as a hardware kind of electrical engineer, who liked internet and software, I wanted to see this technology go further.
That’s a trend I’ve been following through my whole career, is that amazing way of new and better technology to come in, and what we’ve been able to do with that. So, in the early 2000s we were just making this transition from bare metal code that ran on these really-really limited micro controllers. We call them controllers because they literally were hand coded control units, that did nothing else but those assembly [inaudible 8:46] that we put in them. In the 2000’s those transitioned into be really little computers, we started to build real-time operating system kernels, we started to have drivers, we were able to build wireless and wireless communication stacks into them. So, I think about this industry of embedded going through waves of compute.
I recently gave a keynote talk, talking about these waves, and the first wave in the pre-2000s, so 1980’s and 90’s was this bare metal very single purpose, the second wave was bringing in communications, was really what drove the increase in capabilities, the more complex software stacks real-time operating system kernels. But a lot of the focus was still around safety around real-time, and then around some of the communications. In the 2010’s we saw another really interesting wave happen when we tried to bring IoT into these devices, the level of complexity and the amount of compute power that we put into the devices just exploded. So, I always talk about the 2010’s as the time of IoT and of the real OS coming into embedded. We went from kernels to really full-fledged operating systems, if you look at what we build at ARM around Mbed OS, we really built a modern complete operating system stack, in order to deal with all the complexity of internet protocols, cryptography, we’d have to implement things like TLS and DTLS on these micro-controllers. Cloud communication stacks, device management stacks, all that had to get squeezed into these devices, which now transition to 32-bit micro controllers based on Cortex-M, in order to handle all that complexity.
So, I think IoT really drove this 2010’s wave of compute in embedded; and what we’re seeing happen now is really exciting because we got the compute, Moore’s law caught up with what we wanted to do, we’re now packing more transistors, more compute power, a lot of math into these devices. What I’m seeing happening now is that we’re starting to take advantage of that compute to do smarter processing, to do more with the data, the sensor data that we can get at the device level, and save battery power, save bandwidth. We do a lot with LPWAN these days, and there’s very little bandwidth there, and that’s a trade-off for range and coverage.
So, I really think the 2020’s are going to be driven by this advanced processing, really using machine learning right at the center edge, taking advantage of the compute power that’s there, because of these previous waves that went through. This is a thread I’ve been following through my career that I think is really fascinating to see how embedded has just improved by orders of magnitude.
I was going to take a sidebar there and talk a little bit about your experience as a founder, but because of what you’ve just relayed I think we’ll jump into Edge Impulse, because I’m very interested to see how this is converging to your newest entrepreneurial activity in that, and then maybe we can sit back to some of your prior ones as well.
I’m happy to, yeah. Edge Impulse is really the result of myself and my Co-founder Jan Jongboom, getting involved with developers at Arm. While we were at Arm, we ran a developer Open Source activity where we would go and work on projects out in the open with developers. We did a lot of their early work around Python and Java Script on embedded, taking advantage of that in new compute I talked about. But we also started looking at what’s next, and one of the things we were hearing from our developer community was that they’re interested in machine learning, but they have no idea how to apply it on a micro controller, it’s a big mess! Really-really complicated stuff.
So, we did this work to collaborate with the developer community and our own development team, to try and make that easier, how could we bring machine learning algorithms down to the micro control level, and really make it a tractable problem for normal embedded developers, not data scientists, not ML experts, just a normal developer that in their day job they work with devices in some way. We ended up creating some pretty cool projects about 2½ years ago, something, called Microtensor that helped merge the big ML models that came from TensorFlow World, down to these more compact quantized optimized models that can run on a microcontroller without special acceleration, and with a tool chain that was understandable for the developer. That was really successful, we got a lot of developer interest, we could see a lot of applications that were possible, but we also saw some problems; it turns out that just getting the math to run on the device isn’t the hard part, the hard part is the whole rest of the life cycle around there now, because you’re dealing with data rather than code. We have to somehow collect that data, organize it, generate models using it, test those models, version those models, test them again before you can go to deployment, and eventually you get to that part where you generate the math that goes on the device.
This lifecycle of ML, especially when you’re dealing with sensors and devices, is super-super complex and it’s very hard to piece together yourselves, regardless of the size of the company. So, that’s what drove us to found Edge Impulse, was to solve that problem of how do we get this lifecycle to work for people that want to apply ML on embedded? And, how do we make it easy for these developers who don’t have data science and ML expertise, to go and get started; collect the sensor data, apply that sensor data with some great off the shelf algorithms that will solve real problems like using vibration for predictive maintenance, whether it’s anomalies or specific failure mode classifications, using audio for detecting faults or potential activities that you’re interested in the logistics, using images to detect anomalies for example in a manufacturing line, or for healthcare.
So, there’s all kinds of applications that are really motivating for all of us, that we want to be able to do on these devices, and our job is to enable that.
Are you seeing any early, call them first use cases, or killer use cases? You’ve already mentioned predictive maintenance or analytics as an interesting one.
Yeah, we’re seeing a few trends that are really driving early adoption. Of course, there’s interest from our developer community all over the place. You name it, if it’s in embedded they want to apply ML. So, we’re seeing that long-term this will go to every market segment, every application where there are 32-bit microcontrollers involved, there’ll be some aspect of that now, but that’s long-term. Short term, we’ve seen a lot of interest from the health industry, and that’s really key right now under this pandemic.
The health industry has been making use of sensors and technology for a long time, but they’ve got everything they can out of sensors. Some of our early lead customers and projects around monitoring of people’s health, and this gets down to the biometric level, the bio signal level, so monitoring ECG, monitoring motion, monitoring temperature. Then driving more advanced information out of that, like how is someone sleeping, how is someone reacting to a potential sickness like COVID-19. So, we’re involved with that a lot, and I think that’s going to drive the early adoption of this strongly, especially during and post-pandemic we’re going to see that become even more common.
We’ve also been helping with other aspects of COVID-19, one of our users is developing a cough algorithm for example, using audio to detect coughing and sneezing, and that’s something that could be deployed in public spaces, or in someone’s home, to just get a feeling for how people are doing, how much activity is there that could be spreading a disease. So that’s something one of our developers is doing on the platform. So, this area of health I expect to grow, and get more advanced.
We’re seeing lots and lots of activities in asset tracking, and asset health, and that includes humans as well. So, in professional commercial applications we are seeing tons of interest in trackers for worker health & safety, and so this is getting into anomalies in the behavior of workers, potential safety situations, someone’s fallen down, or a tool has malfunctioned right next to them, and that could be a potential safety situation. So, getting help to workers faster, and right when it’s needed. As well as managing valuable assets, is an asset being used? How is it being used? How often is it being used? Some of those are really hard to tell the difference between a machine that’s just parked, and a machine that’s still but in use, how do you tell that difference comes up quite a lot.
And then of course, predictive maintenance. I think predictive maintenance in industry is huge, and it can be applied over so many different things, over so many different sensors. I’ve had the pleasure of being an advisor for the company Petasense, which is a San Jose based start-up in that space. It’s just been fascinating to see where the technology has gone, what MEM sensors have enabled us to do, very cheap, very high-precision sensors that we can put right on these machines, and even after the fact, we don’t have to build them when the machines are made, we can just stick on a small wireless battery-powered device, and make these kinds of measurements.
You’ve talked a lot about the developer community, and it feels as a route to market in some sense, and I know you did a lot of work in Arm around this as well. I guess as you look to develop your ecosystem, who do you look at it in terms of who has done it well out there, how do you model best practices for developer ecosystem development?
The developer community, and community in general, it’s really powerful when you’re working with software. There’s two different ways to go about a go-to-market with software technology, one is the enterprise route, where you go enterprise first and you really only care about the enterprise. Then you care about the developers when you’ve landed an enterprise deal, then you start working with the developers and helping them out etc., and that’s just about product support. That’s the path that we took at Sensinode as well, we tried to make our technology available as widely as possible, but really, we had an enterprise model.
What I’ve learned since then and working with developers, is that actually the developer and the developer community can be a really powerful amplifier to your message, to extending your product and educating the engineering community when something is very new. So especially when you’re introducing new technology there’s a rate of change that an industry is able to move at, and it requires these early adopters and developers who want to learn about it, before there are applications in their workplace, in their day job, they do need to learn about these things ahead of time. Machine learning happens to be one of those that right now developers want to learn about, in fact we found from our research that 40 percent of all developers right now are interested to learn more about that now, and that includes IoT developers.
So, developers are a way to go and see the market, bring in new technology, and get advocates for what you’re doing, and eventually developer communities are great lead generation machines. Those developers are happy with what you do, they want to use that in their work when they get a chance to use this machine learning. In the case of Edge Impulse, they’ll definitely need Edge Impulse if they’ve been happy with the experience they get as a developer, for using the platform. So, that’s how I think about developers, the tricky thing is that working with these developer communities is completely different than working with enterprises, so you really have to think about the culture, the messaging, the business model, the way you give access to the technology, the way you allow your users to contribute to your technology, all those things have to be rethought, if you have a developer first approach to go to market.
That is what we’ve done in this Edge Impulse start-up, we’ve used developers as our go-to-market, which is very common in enterprise SaaS, most enterprise SaaS companies, opensource databases, middleware, you name it, tend to have a developer first kind of strategy. People that do this well, I would name folks from that industry, Neo4j and the developer space have done a great job. Of course, when you work with people in DevOps, like GateLab they’ve done a great job with developer community, and in the data space I would say Treasure Data has done an amazing job, which was recently acquired by ARM. Treasure Data built an incredible community around logs and log collection called Fluentd in the fluent community, now part of the Linux Foundation. They’ve built a huge following for collecting data, because they’ve put these great open sourced tools out there and built a community around it. It’s really free traction for what they were doing, with very little economic cost for them, it just took them to put the energy and do it, and do the right thing, and let some other engineers focus on that community.
So, there’s a lot of good examples where that works, but it doesn’t work for everything, so what we see a lot in engineering and embedded, is a lot of enterprise first go-to-markets, and that sometimes just makes sense, it might be so specialized what you’re doing, it might not be appropriate for individual developers to get their hands on it, and help you bring it to market. But software and machine learning is definitely a space where developers are super-important.
Our most recent connection with you I believe came via the LoRa Alliance, another great community in that regard, and I think it was a TTN meeting that you spoke at earlier this year. When you think about low power wide area net width, whether LoRa or license spectrum if you will alternatives, it does seem like an interesting convergence of that which you’re talking about, in terms of Edge Intelligence and of course powered by the cloud, we have an investment thesis that becomes very close to the old MIT concept of Smartdust from the late nineties, or early nineties actually technically, and we invest in companies right around that whole thing. What’s interesting I think in what you guys are doing is the convergence there of this Edge intelligence in the communications you mentioned, in terms of the waves if you will. How do you see these things kind of converging going forward, especially around communications of these edge devices?
That’s a great question. I totally agree there’s a convergence happening between LPWAN and Edge Intelligence, whether it’s machine learning or other techniques it’s really important that we get more out of the sensors and the data that we get right at the edge, and that we can take advantage of the LPWAN technology.
What’s happened is a kind of inverse relationship between compute and radio, this is something that we saw a couple of years ago when I was at ARM, was that Mores Law had caught up with compute, we were packing more transistors and we can do a lot more compute per watt. But radio really hasn’t improved, with radio we’re fighting physics, and what happens there is that we always have a trade-off between coverage, range, resilience, and bandwidth, and the power that you use to achieve that. In the early days in the 2000s we thought that radios were going to become faster and faster, and faster, we were going to have this continual increase in the bandwidth we had available, it turns out that’s not the case. We took a different route and started to appreciate coverage, we want coverage for these devices, and IoT in particular when these devices are moving everywhere, we made a big trade off around the amount of bandwidth that we have available.
So, what’s happening is that this is a magic combination, you have very limited bandwidth, very low power devices that are expected to live for sometimes years on battery, but you have very advanced sensors and compute available right on these anti-devices. That’s where ML becomes a must have, and we’re going to see that more and more, and then becomes a must have because of battery, or bandwidth, or cost constraints right at the sensors or at the edge. So, when we start combining them now with LoRaWAN, we’re seeing that every single time when there’s a more advanced sensor, so there’s more data than the LoRaWAN network can handle, that’s a place where we can apply machine learning. So, we’re seeing a lot of business cases around LoRaWAN in particular. And I totally agree, the LoRaWAN community as a developer community, as a grass roots movement is just amazing, that’s another reason that we love to collaborate in that eco-system.
The model with its flexible deployment certainly is much closer to the ethos of the community if you will of development open source kind of crowdsource if you will, because you don’t require an operator to be in the middle, which we’ve always liked about it as well. I think ultimately it will help power that.
Let’s switch topics a little bit, and you putting on your investor hat, I’m fascinated with your Angel investing, you’ve already mentioned one of your companies. What are some of the traits of a company that you look for prior to investing to support your decisions?
The investing is interesting, like when I first exited from ARM and I had the ability to go and invest, I’m not one for the stock market, I don’t really like that kind of risk without doing the work, if you know what I mean? I like to be able to influence what I’m successful in, rather than just being lucky! So, I wasn’t really interested in that, but I wasn’t confident that I’m the right person to invest in start-ups, why me, why can I help? I’m a young engineer who happened to build a company and sell it. So, I was a little unsure at first, it was interesting, it was actually a talk that I listened to when I got an award from the Nokia Foundation back in those days for doing IoT work. Jorma Ollila spoke at this award ceremony, and Jorma Ollila was the CEO of Nokia back in the boom days when Nokia was really growing as a mobile phone company.
Something that he said which really struck me, he had just donated €5 million of his own money to the Nokia Foundation, that’s a Foundation that provides grants for graduate students studying in technology, they want to ensure that there’s new technology out there, and graduate students can go and do their research. So, Jorma contributed €5 million of his own money to this, and his reasoning was really simple, when he was a young man and a young engineer, he was funded by a scholarship from a foundation in Finland to go and study at Oxford in the UK. That changed his career, that was the thing that enabled him to go on his trajectory and become the CEO at Nokia and make that successful. So, his reasoning was simple, ‘This is what made me successful, if I can help do the same thing for your people that’s going to make them successful, and that’s good use of my money and my time.
I went, ‘Wow!’ I actually really care about bringing new technology to people about building start-ups, and I have some experience, I’ve done this, and I can help others. So that was the moment I decided, ‘Alright, I’m going to dedicate any money that I make from my start-up activities into other start-ups and try to do some mentoring for other entrepreneurs. That was the ‘Ah-ha!’ moment for me, and so I founded my own venture investment fund that I make investments out of, and it’s like a small greenfield fund. What I look for is really simple, I’ve always appreciated vertical go-to-market strategies when you can bring a technology much closer to the end customer, and I’ve always wanted to be involved with those. But at the same time, I’m a deep-tech kind of person and leader, I’m able to bring deep-tech to the embedded developers, and it’s hard to do both at the same time!
So, I’ve always wanted to get involved with verticals, and this was a chance for me to do that. So, the investments I’ve made have been much closer to the verticals, and typically focused around commercial and enterprise industrial applications where I know a lot. So, for example Valka lighting that I’ve invested in, is a leading commercial industrial lighting system provider in Scandinavia, IKEA for example is a customer in Sweden, they provide the entire lighting system, the lights , the control systems, I’m a control wireless system guy so that was exciting. Augumenta which does industrial augmented reality, how could we bring augmented reality into industrial settings? Really-really important right now with what we’re seeing with the Coronavirus, and the lack of manpower available in some facilities, this helps with that. So, getting right down into the user interface of compute in industrial environments is exciting.
I got involved with computer vision and machine-learning in real estate, so I’m an investor in CubiCasa, and CubiCasa is a leading virtualization collection tool, uses a mobile phone to take a video of a piece of real estate from the inside, an apartment or a home, and it uses computer vision and machine learning to turn that into a floor plan that the real estate agent uses in the listing. That’s an amazing piece of tech to just digitize interior spaces, and you can use this for real estate, super-important now during pandemic when people can’t go and physically visit every place before they’re serious enough to go and take a look. And so, investments like this have driven me to where there’s a path to market in industry, we can influence that.
The other thing I look for is the use of machine learning and computer vision, I really think that’s going to make a huge difference in the future, where we can apply this technology, so I want to see this technology be applied at what it can do. So, I always look for that angle. And then of course, just a great founder team, founders that I understand and that I think I can help, because if I can’t help then my money doesn’t matter, it’s not about the money from Angels, it’s about their help and their network.
Truly smart in that regard. Fascinating, because our Fund One which is our own Evergreen fund where the LPs for that probably matches pretty closely your both thesis and focus here. So, post-this we’ll have to certainly compare notes on some of those, because it sounds like fascinating companies.
One final topic, and we’d be doing a disservice if we didn’t talk about this a little bit, is the work that you’ve done around the Micro:bit Foundation, and how it gets kids interested in technology, probably really relevant these days as many of us are sheltering in place with our pre-teens! So, tell me about what you’ve done with the Micro:bit Foundation, your inspiration and a little overview of it.
So the Micro:bit was an incredible project that was started by the BBC, where the BBC looked at the previous ways of compute, and like we had the Commodore 64 in the US, the UK had the BBC Micro. Interestingly enough, the BBC Micro is how ARM got started, Accor Computers was part of that, and eventually ARM spun off as a processor design, that was made with the first BBC Micro.
So, that had very strong history in the UK and the BBC wanted to look at, ‘Well, what’s next? How do we improve young people’s interest in technology? How do we improve their interest in coding, and STEM? And most importantly, how do we get more girls involved in technology?’ because there was a huge divide in the UK particular, between girls and boys and their interest in STEM and technology in general. The interest levels are very low for girls, so, how can we fix that?
The BBC did something really unique which was they got people who weren’t engineers involved with building this thing, so psychologists, teachers, education experts, branding experts, people that rethought what the experience might be for young kids getting involved with technology. Like, hat’s the Commodore 64 moment for kinds today, that’s the way I think about it. That project ended up creating a little embedded computer, it’s about the size of a credit card, it has just two buttons, 25 LEDs that make up a 5 x 5 matrix, some sensors like motion sensors, and a Bluetooth low energy wireless interface. It’s a little computer that can run programs that the kids create and allows you to wire stuff up to it, like Servos, LEDs, other external sensors send messages between the machines. So, it’s really reinventing this, ‘I can create’ experience for young kids today.
The BBC did a brilliant job doing that, but of course to make that happen they had to collaborate very broadly across the industry, they don’t have the capabilities to bring that kind of product to market. So they brought on ARM, Microsoft, Samsung, and many-many other organizations to go build a project to bring this Micro:bit and give it away to every single Grade 7 student in the UK. This is back in 2016 that the giveaway happened. That was the first time a computing element like this was ever given to every single child at a certain grade level in a whole country, so it was an amazing project.
That project was successful, but the problem is projects don’t tend to have a very long shelf-life, projects get wrapped up and stopped, and that’s what was happening after the original project. The team around the original Micro:bit project was looking for help to go and turn that project, and that idea, into a long-term foundation that could bring this technology to kids all over the world, really productize the BBC Micro:bit into something that was long-lasting. I got involved with it through ARM, our team was helping sponsor the project, we were giving engineering resources, one of our guys did the first layout of the PCB, another one of our software engineers did all the original software work as a volunteer activity from ARM.
So when they started looking for help to turn this into a real organization, they realized they were missing this entrepreneurial spirit, they didn’t have anybody that knew how to run and build a start-up that could bring this kind of technology to market in a cheap way. So, I was given the opportunity to take on the cofounder and chief exec role by ARM, it was ARMs exec team, Simon Segars, Mike Muller that gave me that opportunity, so I’m always very thankful to them to have been able to go and work on the Micro:bit.
When we were given that chance, together with my co-founder and CTO Johnny Austin, who is still at the Foundation, we were kind of on our own. Once we got out there we literally went and created that tech start-up as a non-profit from scratch, because there wasn’t anything to start with, so we did things like working with the lawyers to create the entity in the UK that was a non-profit, doing all the IPR work to bring the IPR into the Foundation, and then productizing, creating an online system to support the teachers and the students, making all that web infrastructure work and scale, ramping up the production of the actual product, so making sure that could be produced at the scale that was needed for the rest of the world.
Then you know Ken, this was the most extreme experience in hyperscale that I’ve ever experienced in my career. When you have the BBC behind something like that and making a big noise internationally about how great this was in the education system in the UK, they had 70 ministers of education lined up from around the world to get this thing, and they wanted it. So, it was probably like what some ventilator or mask manufacturers at the moment are feeling, where ‘Oh my God, every hospital in the world now wants what we have, and we don’t know how to scale this thing’. So, we had to figure that out very quickly, and so in a period of about 18 months we went from nothing to 50 countries in deployment, it was just how do you get out of the way? How do you use ecosystem, how do you use partners to go and do this thing in every country? How do you get out of the way of manufacturing, find ways to fund manufacturing at massive scales where the small cash reserves of a small non-profit are in the way?
We had to figure that stuff out as a start-up team, and the team did an amazing job to go do that, it really was a case of hyperscale. That was an amazing journey to go build that, I eventually had to go back to California, I came back to Europe for two years to do that, I had to go back to California, and we decided to take on an operational CEO at that point who could just run and operate the foundation. Things were already scaled, it was profitable by that point, so it could run on its own, and I went back to ARM to work on these developer things that eventually ended up becoming Edge Impulse.
That’s quite a story. If there’s a single word that I think so far has described every element of what you’ve talked about, for me its impact investor, truly impactful if you will and certainly as an entrepreneur.
Final question that we always like to ask on these podcasts is, just general recommendations of books or resources that you’d like to share. What inspires you?
Like a lot of engineers, I don’t read a lot of non-fiction, because I deal with so much non-fiction in my day job, a hell of a lot to do with. Sometimes you need to escape and imagine a little bit to open up your thinking, and so I read science fiction. Science fiction is a way to think about things differently and think about the future. A book that I’m going to recommend which opened my eyes to a lot of fun things, and made me laugh out loud many times, my wife would look at me strangely when I’m laughing out loud reading a science fiction book! This book is called ‘The Punch Escrow’ by Tal M. Klein. It’s particularly appropriate right now because it talks a lot about artificial intelligence, and 3D printing at a nano-scale, as well as self-driving cars. There’s a fun point in the book where humans are no longer allowed to drive, self-driving cars are truly only self-driving, and its only emergency situations that a human can hijack a car and drive. There’s one point in the book where a man was injured and had to steal an ambulance, drive this ambulance, and there were alarm bells going off in the network of AI self-driving vehicles going, ‘Oh, my God, there’s a monkey driving this car, we’re all in danger!’ as responsible self-driving AI cars. ‘Everyone get out of the way!’ panic-panic-panic. It sounds a lot like today where we’re worried about the one self-driving car on the road, in the middle of all these people driving. So, things like that are always eye-opening, for what the future may hold and stuff.
Great recommendation, I will definitely have to put it on my read list. So, Zach it has been a true pleasure, I really sign off saying Zach is an impact entrepreneur, impact investor, and impact technologist in embedded space, we’ve been very pleased to feature you on our Digital Leadership Podcast Series. Zach, thank you so much for taking the time and sharing your brilliant and impactful insights.
Thank you, Ken, it’s a real pleasure.
Alright, thank you.