Founders on Fire: Nick Romano, CEO and Co-Founder, Deeplite Founders on Fire Podcasts Posted by Jon Howell | 23/08/2022 Today we’re catching up with Deeplite, our winner of the 2021 AI Trailblazers Award. We have the pleasure of chatting with Nick Romano, CEO and Co-Founder of the firm. Chief Trailblazer Rose Ross quizzes him to find out more about how Deeplite came about, and what exciting work the company is currently involved in. Nick shares how the Waterloo and Montreal corridor is a powerhouse of AI in Canada, how Deeplite is helping BlackBerry’s IVY platform in allowing third-party AI-based synthetic sensors to obtain and process data from cars and trucks in real time, and how their DeepliteRT purpose-built inference engine is making AI possible on legacy systems which would typically require new, more powerful hardware. He also advocates the importance of remembering that a startup should be treated a family, including the families of those involved. Listen to the full podcast here: YouTube: Also available on: SpotifyAnchor Interview transcript Rose Ross: Hello everybody, and welcome to the Founders on Fire. My name is Rose Ross, and I am the Founder and Chief Trailblazer at the Tech Trailblazers Awards. I’m delighted to be joined today by Nick Romano, who is CEO and Co-Founder of Deeplite, our AI Trailblazers. Hello Nick, how are you doing? Nick Romano: I’m well, how are you? Thank you for having me. Rose Ross: A pleasure. So, it’s been a number of months since you were crowned the AI Trailblazer, so I thought it’d be really nice just to do a little bit of a background on you Nick, and how you and your co-founder got to the position of, in 2018, founding Deeplite. You’re based in Canada, Montreal, so it’s really nice to say a nice maple-leaf-based ‘hello’ to you as well. Nick Romano: Perfect, perfect. I can give you a little bit of the genesis of Deeplite. Maybe I’ll start with a little bit about my background first. I’m an engineer by education. I’ve been in the tech space since the late 90s actually, so I’ve been around for a while. Three-time founder, before Deeplite I was co-founder of an enterprise SaaS company, based in Toronto, Canada. I exited that through a private equity transaction back in 2017, and shortly thereafter I was introduced to a group in Montreal, called TandemLaunch, it’s an incubator programme in Montreal. They introduced me to Ehsan Saboori and Davis Sawyer, who were the original co-founders of Deeplite, who were working on, at the time basically, the Deeplite project within the TandemLaunch incubator. So, I got to know those guys pretty well over a few months, really loved the story of Deeplite, and to make a long story short I joined them as the third co-founder, invested some of my own capital from my last exit into the business, and then we graduated out of TandemLaunch into a standalone business in 2019. Rose Ross: And do you want to tell us a little bit about Deeplite itself. You talked about the Deeplite project, what is the AI that you are bringing to the world? Nick Romano: Our tagline is ‘AI for everyday life.’ Our whole core focus, and our core thesis is about enabling AI in the things that we use every day. We fundamentally believe that the edge is going to be bigger than the cloud in terms of the opportunity to deploy AI. One of the biggest challenges with doing that is AI is a very compute intensive process. It’s algorithmic, it requires a lot of processing power, a lot of memory, a lot of just sheer power, electricity to run AI. So, if you’re battery operated for example, it can be problematic. So what Deeplite does and what we’ve created is, we’ve created an algorithm and a platform where we can take an AI deep neural network, that say it’s been trained for accuracy, so it does its job in terms of what the user is hoping for it to do. But maybe it’s too big, it’s too slow, too power consumptive, and they can use our tools to turn that AI into a much smaller, faster, less power consumptive version of its former self, without compromising that original accuracy. This way you can now deploy it into more constrained environments and processors. Rose Ross: So you’re basically slimming down the devices that are AI warranted. Nick Romano: Yeah, exactly. So rather than having a car with a bunch of computers in the trunk, in order to run all the AI, you can deploy it on much smaller, economical and practical hardware. Rose Ross: And you’ve done a fantastic segue for me there Nick, without realising it. Because I noticed that you guys have recently entered into a new programme which is supported by BlackBerry, which is very much focused on the automotive, autonomous vehicles, world. Nick Romano: Yes. Rose Ross: Can you tell us a little bit about that? So that’s a programme with I believe another four or five organisations, as I said sponsored by BlackBerry. Can you tell us a little bit about what’s the name of the programme, what’s it about, and why is it important for you and your cohort? Nick Romano: It’s actually a really awesome programme put on by another great Canadian company, BlackBerry. And some folks who are familiar with the BlackBerry name, don’t necessarily realise that a big part of BlackBerry today is an operating system called QNX, and this real-time operating system is deployed in literally over 200 million vehicles globally. So, if you’re driving your vehicle, your infotainment system and other controls within the vehicle, they’re all running on this operating system. So, most people don’t even realise that there’s BlackBerry in a lot of the vehicles that are out there. And so BlackBerry has launched a new programme, a new opportunity in the automotive space called BlackBerry IVY. And what IVY is, is it’s leveraging the fact that BlackBerry is in the vehicle and has access to all of this rich sensor data within the vehicle, whether it’s climate, infotainment, drive train, all of this information. I mean, cars are basically giant computers these days, and essentially what IVY is all about is about getting access to all of that sensor data, making it available, so that third parties can create what they call synthetic sensors, or in layman’s terms you could call them essentially apps, so that third parties can create apps based off of getting access to all of that wonderful data. So, what we’re doing as part of the programme is a really good alignment to Deeplite’s solutions, and how our solutions have evolved from an AI implementation perspective. So we’ve got our wonderful solution stack for making AI practical and accessible, and making it easier to deploy. So, what we’re doing is, we’re embedding our tools and demonstrating how our tools can be a fundamental part of the IVY platform, so that third parties who are building these synthetic sensors, if they’re AI-based synthetic sensors, they can use our tools as part of the journey to create those synthetic sensors. So, it’s a pretty exciting collaboration with BlackBerry, it’s put on by another organisation in Ottawa called L-SPARK, they’re the ones that host this programme and brought BlackBerry to the table, and then selected Deeplite amongst only four. There’s only four companies, it was originally six, they decided to make it even smaller just because of the focus that they wanted to put on the programme. Rose Ross: Fantastic. And that is based in Montreal? Because it is in Canada, but I wasn’t quite sure if it is in Montreal. Nick Romano: Yes, it is in Canada. The BlackBerry organisation is the spread between Ottawa, Waterloo, and Toronto, so it’s primarily in Ontario, but Deeplite has an office in Toronto as well, and that’s actually where I’m based. So, it works pretty well. The corridor between Waterloo and Montreal, it’s about a six hour drive from end to end. So, it’s not too far. Rose Ross: Sounds like a great road trip, but not something you necessarily want to do on a daily basis. Nick Romano: No, definitely not, definitely not. Rose Ross: Are you guys physically getting together then with the BlackBerry team, with the catalyst of L-SPARK, to help bring everything together? And are you starting to collaborate with your cohort organisations as well, is there collaboration possible in that way, too? Nick Romano: Yeah exactly. That’s one of the cool things about this programme is, given the various locations of individuals, it’s still primarily virtual, but we definitely get together I would say, at least monthly we’re getting together either in Ottawa, Toronto, or what have you. And one of the other really cool benefits of the programme is interacting with other cohort partners. So, we’ve got this AI system that we’re putting in. Actually, I have a device in my hand here from one of the other cohort partners, Raven, who has this device that they can plug into the vehicle, and it’s got an in-cabin camera and outside-facing camera. They can do some pretty cool things with their device on using some of that sensor data. So they’re creating essentially an app version of that device I was just holding, on the IVY platform, and we’re talking about using our AI tools to help enable some of the features within the Raven platform as well. And there’s other cohort partners that we’re starting to explore ways we can work together as well. So, we’re creating a little mini ecosystem here around this, which is great. Rose Ross: That’s exactly what people need isn’t it, to have the innovation and specialists pull those things together. But, let’s go beyond autonomous cars or vehicles technology, generally speaking, because really you don’t care do you? Nick Romano: Right, no we don’t, exactly. Rose Ross: With some AI we can make it better, right? Nick Romano: That’s exactly right. Rose Ross: So, what other areas are you working in? I’m sure there are some other hotbeds of requirements for this type of AI technology, so can you give us any real world examples, or even stuff that is a bit more in the pilot and visionary stage at the moment? Nick Romano: We’ve worked on some pretty interesting projects, and I think that’s one of the cool things about what we do is, just hearing use cases that we’d never even thought of, where people are saying, ‘Hey, we’re trying to do X-Y-Z with AI, can you help us?’ So, in consumer electronics, for example, we’re working with a very large organisation on a smart fridge, where we’re using computer vision to detect what’s going in and out of the fridge. And of course refrigerators don’t have a tremendous amount of compute horsepower, so it’s a very, very constrained environment, they want to have that capability in the device itself, not in the cloud, so where you’d be sending sensitive data back and forth. So, that’s one example. In manufacturing… Rose Ross: But do the also get people to lock the fridge after about nine o’clock at night? That would be quite good! Nick Romano: That sounds like something that could be a feature for sure, or at least an alarm of some sort. We’re working in industrial manufacturing, we’re working with a steel company in Japan that’s doing anomaly detection on the rolled steel that they’re creating. And also there’s a handheld device that we’re working on with them, they etch the metal when they create a new batch which has information about the steel. So quality inspectors will use this handheld inspector device to read that information and interpret that information. That’s also AI, that’s an embedded device that’s running on an Android operating system, it’s battery operated, it has to be very efficient. We’ve got a really cool customer, another Canadian company called E-SMART, where they’ve created this device which is a speed control system for trucks, so they’ll sell into fleets and things like that. So basically, a fleet manager can say, ‘We’re not going to let our trucks go more than 10 miles an hour over the speed limit,’ or something like that. And the original version was GPS based, so you use maps to determine the speed limits on given highways or what have you. But the big challenge for them was, if there was a construction zone, or some other dynamically changing speed limit that was temporary, it wasn’t reflected on the maps, trucks would blow through these construction zones and they’d be fined, etc. So, they’re using our technology to enable a computer vision solution to read the actual speed limits in real time as the truck is traversing the highway. So it’ll read the posted sign and say, ‘Well, no, there’s a mismatch, so you’ve got to slow down to 40 miles an hour,’ or whatever the speed limit is, and they’ve got thousands of these units in the field. In order to enable this capability, they’re using Deeplite’s technology so that we can fit this AI into the existing footprint of their hardware. Which is a bit of a theme for us, we’re starting to see this in a number of areas where rather than replace all of this hardware and try to source new semiconductors and all of that stuff, which comes with its own logistical and cost challenges, we can now fit the AI into the existing infrastructure, breathing new life into products that might have been at the end of their shelf life. So, there’s pretty solid economic impact for the customer, but also a good eco story there, in terms of being able to reuse and repurpose billions of chips that are already out there in the field. Rose Ross: Well, that’s going to be incredibly important, isn’t it? We know there’s a lot of pressure on the supply chain for chip manufacturers, sourcing raw materials to go into them, etc, etc. I think we’d all like a little bit of that recycle-reuse type of viewpoint to go into tech, so you don’t completely chuck everything away and start from scratch. Nick Romano: Exactly. Rose Ross: That’s really exciting. Tell me a little bit more about what else you’ve been up to, or what you’ve got planned in 2022. Nick Romano: So, where we’ve historically been hardware agnostic in our optimisation capabilities, we’ve launched a new runtime engine called DeepliteRT, which is a purpose-built inference engine, which runs the AI on the device. The early, original generations of our technology were very much in the AI/ML development pipeline. So you train a model, you’d optimise the model, and then you’d deploy it, compile it for whatever hardware you’re running to. So we would basically hand off optimised models to other software stacks for different types of hardware. We’ve extended our solution now to encompass the actual inference on the device itself. So, we’ve created a very, very efficient inference engine, specifically designed for commodity hardware. So, we can take these efficient models and make them even smaller, with some novel methods for what’s called quantisation, and we can make them even smaller, even faster. Rose Ross: Quantisation, what’s this? You’ve just dropped in a little bump word I’ve not heard before, so come on then. Nick Romano: I try not to use that word too often, because it’s very labby! But it’s an algorithm where you can essentially transform, in this case, a deep neural network, down from what’s originally a 32-bit network, down to what’s called lower precision. Typically, what’s very common today with quantisation is taking it down to 8-bit. So essentially you can get a four times reduction in size, when you go down to 8-bit, but what Deeplite is doing is we’re taking it down to 2-bit and 1-bit, which makes them really, really small, and really, really fast. And again, the trick for us is we can do that without affecting the accuracy, because typically the lower the precision that comes at a cost, and that cost is typically the accuracy of your model. So, with this new direction for Deeplite and this new capability, we can quantise these models down to 2-bit, 1-bit precision, make them super-small, super-fast, preserve the accuracy. But there is nothing out there in the market that could run the 2-bit network, so we created DeepliteRT for that purpose, so our runtime engine can run these mixed precision networks that we create. Rose Ross: Oh, wow. Nick Romano: Yeah, so that allows us to go into even smaller hardware. Right now, we’re into small ARM Cortex-A type CPUs. Rose Ross: And how big are we talking there, just to help me visualise how big that is now? Nick Romano: Very tiny, your Cortex-A chip is very, very small, you can put that into pretty small devices. We’re also now looking at microcontrollers and sensors as well, which are even smaller. These things would be embedded in an electric toothbrush and stuff like that. Rose Ross: Interesting. It’s good to check on the kids with that, hey. ‘Got to go for another three minutes.’ Nick Romano: Exactly. Rose Ross: Obviously, for you guys it’s been important to get the recognition, you’ve had a stamp of approval from ourselves last year. Obviously, the stuff that you’re doing with BlackBerry is really important for your growth. What other things that have been important to you, personally perhaps Nick, on your founder or co-founder journey, because it’s not easy is it? You sound like an upbeat positive type of guy, and very smiley, he’s gonna have a little grin there, because you don’t know what I’ve just dealt with this morning. But it’s not just a little buttercup-lined path, is it? It can have its challenges as well. Nick Romano: For sure, yeah. As I mentioned at the beginning, this isn’t my first rodeo either, I’ve been around the bases a few times, that experience has helped significantly. Listen, when somebody is deciding that they want to follow this path, it definitely comes with its own set of challenges, much like a lot of career choices, and it’s a commitment. And one area that I think is often overlooked is that it’s not just a personal commitment, you’re making a commitment if you’ve got a family, you’re making these decisions on behalf of a group of people, because it affects more than just the immediate employees, founders, and team. I really like to look at a company as a family from that perspective, and it’s the extended group that is part of the journey. It’s not just the pay-rolled employees and staff that are part of it. I think decision making as a leader in a company, you need to consider that people are people, there’s a human element to what we’re doing. It’s not just about business, it’s not just about the numbers. I mean ultimately at the end of the day for the survivability and sustainability of the business, it is, and tough choices need to be made, but they need to be made from looking through that kind of a lens, and just keeping the humanity part of the story. And at the end of the day, to me that adds loyalty, brings loyalty to the business, respect to the business, and just helps with keeping everybody rowing in the same direction, and feeling like they’re part of something. And that’s the key for me. Rose Ross: It’s an interesting thing, because I know with your accent people probably assume you’re American a lot of the time. Nick Romano: That’s true, yes. Rose Ross: So when they talk to you, they go, ‘Oh yeah, you’re from America?’ Well, kind of very North America! The Canadian ecosystem, I don’t have an awful lot of exposure to that, obviously a lot more to the US VC-driven, very fast paced on the west coast, on the east coast a bit more conservative and metrics oriented, etc, etc. How do you feel the Canadian tech startup ecosystem is doing? I’m not saying it’s in the shadow of your neighbours, because I think that’s unfair, because you’re very different. But in relation to that, do you feel that you’re making enough of a splash, so to speak, bearing in mind all the stuff that’s going on there at the moment? Nick Romano: I’m actually a huge advocate for the Canadian tech ecosystem. It’s quite robust, and increasingly becoming very integrated from a North American perspective. Actually more than 50% of our investor money has come from the US in Deeplite. And that corridor, interestingly enough I was mentioning before, it’s the Waterloo, which is west of Toronto, through to Montreal, that corridor there, it’s an incredibly robust AI hub for North America. There’s a lot of universities, there’s a lot of AI skills, and people with that background in the Canadian ecosystem. So we’re starting to get that reputation as being a Centre of Excellence in North America for AI in particular, not just limited to AI, but that’s just one area where we get a lot of attention. And what comes with that is, you’ve got a really robust startup community, but we also have some of the giants from the US, that have set up some pretty big operations in Montreal, Toronto, Ottawa, and those locations, which makes the recruiting a bit of a challenge when you’re a startup competing with some of the some of the bigger software operations that are out there. But, at the end of the day, it all adds to the ecosystem. And then even employees in these big shops who maybe come out of university, go get a job for a couple of years, they’re the next generation of entrepreneurs that will walk out of those doors and set up a new startup in Canada. So, it all works in the end, I think. Rose Ross: Exciting. It sounds like you’re having quite a buzzing time there with all the stuff that you’re up to. Anything else that you’d like to share, either from a personal experience, or from a Deeplite perspective, or even what’s happening in AI generally? Nick Romano: There’s a lot of entrepreneurs out there right now that are experiencing an economic situation, a challenging situation geopolitically, economically, where the venture community is stepping back a little bit. It’s not quite as easy to raise money as it was maybe even a year ago. So, for those entrepreneurs who haven’t seen those cycles, or haven’t experienced those cycles, I’ve been through the ‘.com boom’, the financial crisis, these things happen and they’re cyclical, and you’ve just got to figure out the best way to survive through some of these downturns, and preserve cash, and do the things that you need to do, because it will turn around. So, if anybody’s listening that’s very worried right now, I understand, it’s tough, but this is part of the journey that we deal with as entrepreneurs. Rose Ross: The natural cycle of business – this too shall pass. Nick Romano: Exactly. Rose Ross: Well unfortunately, this interview shall also pass! We’re just winding things up, but it’s been fantastic, very insightful, Nick. Nick Romano: Thank you so much, it’s a pleasure. Rose Ross: Fantastic. Well, that was Nick Romano who is the co-founder and CEO for Deeplite. He has been sharing with us some of his experiences here on the Founders on Fire podcast. That’s brought to you by me, Rose Ross, and the Tech Trailblazers Awards itself. You can find out more about us at Techtrailblazers.com, on Twitter @TechTrailblaze, and you can also find us on LinkedIn. Thanks very much everyone. Nick Romano: Thank you.