On the Record with Nick Romano, CEO at Deeplite
Hometown: Toronto, Canada
Favorite video game: Ace Combat
3 words to describe yourself: funny, driven, and loving
Deeplite is enabling AI for everyday life. We use AI to automatically make other AI models smaller, faster, and more energy-efficient creating highly compact, high-performance deep neural networks that can run at the “edge” in vehicles, cameras, sensors, drones, phones and many other devices you use every day.
What were you like as a child, and do you think any of your childhood experiences influenced your decision to become an entrepreneur?
My parents gave me a lot of latitude. Good behavior paid off, and they gave me a long rope to do the things I wanted to do. I was also always very busy, and in my teens I joined the Royal Canadian Air Cadets. The group was affiliated with the Canadian Air Force and I developed a lot of important skills around leadership, teamwork, and goal setting. I also got my pilot’s license through that program, and since it was the late 80s when Top Gun came out, I was a pretty bad ass 17 year old in my flight suit.
I know your parents were born in Italy and moved to Canada before you were born. How do you think this affected your upbringing?
My parents immigrated to Canada from Italy after World War II, but they were young when they came over and met and were married here, so my childhood was a real combination of that old world mentality with the transition of becoming North American. I grew up with the traditional, European, family oriented mindset, but it was really important to my dad, in particular, that I fit in. He was bullied as a kid because of his immigrant roots and so he resisted the urge to raise me in more Italian ways. For example, he would only speak to me in English at home.
How did you get into entrepreneurship
My dad was an entrepreneur his whole life, and as I watched him grow his business—I really saw him as a mentor. He was very successful in the construction trade, and I worked for him on weekends and in the summers—he really instilled a strong work ethic in me. Working with him taught me about both the opportunities and the challenges of entrepreneurship. Even though it’s a very different area than what I’m in now, I knew having my own business was something I wanted to do.
Can you describe Deeplite in layman’s terms?
In a nutshell, we’ve created an engine—an algorithm—that can take a deep neural network and make it smaller, faster, and more energy efficient. For those that don’t know, AI is typically computationally intensive. It requires a lot of compute power, a lot of memory, and a lot of energy to actually run the AI. What artificial intelligence ultimately does is make rapid decisions on your behalf or on behalf of whatever task is required. For it to do that at the speed we expect, it requires a tremendous amount of compute. That is fine in the lab, but if it’s too big or too slow or too power consumptive, it’s hard to translate into the world. That’s where Deeplite comes in. We want to make AI ubiquitous and decentralized in the things we use every day. Self driving cars is an obvious example, but we’re also talking about the AI in an electric toothbrush. We’re working with a large coffee maker company that is making a smart coffee maker for pods. You’re going to start seeing AI deployed into mundane things. People don’t think about where AI is used—our phones, for example—but in order to make this happen, and to truly unlock the potential for everyday use, we have to fit AI into the world and not the other way around.
We know you’re working on a secret project for a large toy company, can you share any information about that?
This toy company called us out of the blue, and we were connected to their computer vision team (sidenote: who knew they had a computer vision team?!). It turns out that they have this creative play lab where they’re working on apps for tablets and phones for a really immersive, play experience. The idea is that with the camera on your phone, you can use the app to turn the product into a digital experience.
In the context of privacy, is there anything you’re worried about in terms of a child’s toy having deep neural network capabilities? Are we ever worried about toys becoming too smart?
Privacy is one of the main value props that we encourage with our tech. If you can get AI out of the cloud and into the edge device itself, then you’ve addressed a very specific risk with privacy. If you’re gathering data, and you have to send it to the cloud to run AI inference and get results, you’ve got data moving back and forth, and it’s susceptible to being intercepted and used. By having the AI process local at the point of data capture in the device itself, you reduce the risk of that data being compromised. You can keep what’s happening within the confines of home because it’s not being transmitted elsewhere.
Another thing is the whole concept of federated learning where bits of learning are being gathered from places at the edge rather than being brought to a centralized location where you’re training off of central information. If you can have snippets happening locally that are only relevant to what’s happening in your network, this also reduces the risk of data being used against your will.
How would model compression and optimization impact the training?
Our ultimate goal would be to have what we’re doing be part of the training process. Right now, that’s aspirational. At the moment, it's a 2 step process: training is what has the impact on accuracy. So the better the training, the more accurate the model. Our current approach is to use an accurate, trained model that is maybe too big or too slow and use that as the teacher model for our algorithm. So when we do our optimization, there is a training loop as part of our process. As we are optimizing, we are going through training exercises ourselves because one of the key strengths of our work is that while we’re optimizing models for speed size or power consumption, we’re not compromising accuracy of the actual model. We can preserve accuracy. If you’re familiar with training in AI, the more complex the process or problem, the longer the training time. If you’re adding an optimization step that incorporates training, you could be doing a training process for a model that can take days, if not longer, depending on use case. A lot of the core IP that we have is how we rapidly converge on the best optimized model in the shortest time possible (which is pretty fast by the way). Part of that looking ahead is how to incorporate some of that stuff into the initial training step and shrink that timeline even further.
What is your favorite aspect of Deeplite?
The people—we have an outstanding team. The crew is resilient, and they’ve shown an ability to adapt to this new reality while continuing to advance the platform and sales opportunities. We joke that, with COVID, we don’t even know how tall some people are because we’ve never stood next to them, but we’ve done some really cool events virtually as a group, like trivia and poker nights. I’ve had multiple businesses over the years, and the way you are successful is by having the right people around you.
Do you have any particular work mottos or philosophies?
I started my first company at 28, and in my early days, I was a lot more serious. Eventually, I came to realize that you only live once, and you have to lighten up a little. It’s not that you don’t take a challenge or problem seriously, but you can approach it with a different attitude, and that makes all the difference.
What do you think is your leadership style?
I try to find the best people and drive a culture of accountability. If you can't trust the people you have to do their specific job, and you feel the need to do it for them, then you don’t have the right person or you’re not the right leader. You need to be able to delegate to people who share the same value of accountability. You need people who take what they’re doing seriously and can get the job done. If you have enough of those people sitting in the right seats, it drives success.
What is your most favorite and least favorite part about being a founder?
Entrepreneurship has always been in my blood so I’ve always been someone who wants to drive, lead, and create something out of nothing. Being the founder or cofounder of a company is about being able to say “I was a part of creating something cool and new.” That’s my favorite part. On the flipside, my least favorite part, is that sometimes being a founder can be a lonely place because you have pressure from multiple directions. You have to make a lot of sacrifices, but at the same time you always have to have a happy face. I have a tendency to bury my feelings, but if you do that too much it can be bad for your mental health—you have to have the right mind set. It’s a high risk, high rewards scenario.
Do you have any advice for potential future founders?
Family support is massively important. A big mistake I see people make is that they don't realize that the decision to be an entrepreneur is not an individual one. It’s a joint one. If you’ve got kids, spouse, etc..., everybody is part of that journey and everybody makes sacrifices. You have to appreciate and recognize that. If you can do that, that’s how families stay together throughout the journey.
Anything else you want to share?
One of the cool aspects of Deeplite is the actual make-up of our Co-founders. We’ve got Davis, who is in his mid-20s. He is extremely smart, energetic, and a domain expert in what we do. Ehsan is an Iranian immigrant with a PhD. He’s the real brains behind the IP. As for me, I’m the 50+ year old veteran who has been there and done that. It’s a really interesting, demographic mix that I think is fairly unique in a lot of ways. We’ve really leveraged everyone’s strengths and maximized what we each bring to the table.
-------
Nick Romano is CEO of Deeplite, a Canadian AI software company dedicated to enabling AI for everyday life. Nick is a serial entrepreneur delivering successful outcomes through leadership, integrity, innovation, and empowerment for over 20 years. Prior to Deeplite, Nick was co-founder and CEO of several successful Canadian tech companies, including Messagepoint, Prinova Software and Prinova Technologies. Nick earned his Bachelor of Engineering and Management in Mechanical Engineering from McMaster University in Ontario, Canada, and was recently honored by McMaster’s Engineering Faculty as one of their Top 150 Distinguished Alumni for their role in shaping Canada and the world.
Related News
