For this Deep Tech Dive, I had the pleasure of sitting down with Nicholas Nadeau, Project Director & General Manager at Halodi Robotics. A Robotics-as-a-Service (RAAS) working on bringing humanoid robots to everyone.
- Welcome Industry 5.0, robots collaborating with humans.
- Don’t go into full scale or presume you are in production too early. Keep your R&D units where they need to be, in R&D.
- Once again, Proof of Work > Proof of Knowledge. If you want to work on robots, build them in reality, not in theory!
- School should no longer be the rule, just an exception.
“If they’re looking for water on Mars, they should send me up there with a bucket of balls and my 7-iron”
This interview was edited and condensed for clarity.
Tell me about yourself and your background?
I am currently the Project Director of Halodi Robotics and General Manager of North America, growing their North American presence headquartered here in Montreal. For school, I started at McGill University, first for mechanical and biomedical engineering, and then went to École de Technologie supérieure (ÉTS) in Montreal.
Originally I wanted to be a surgeon, but I fell in love with engineering. I used to build race cars and ride them around the world at formula SAE competitions. During my stint as a biomedical R&D engineer, I built neurosurgical devices, brain implants, surgical robotics, deep brain stimulation, and brain imaging equipment, etc.
How did your experiences bring you to Halodi Robotics?
I lived in the medical space for almost 10 years. In parallel, I started a consulting company while working on a Ph.D. in robotics. I focused on collaborative robotics and machine learning (ML), to make hardware smarter. My whole career since then has been on the edge of hardware and software at all points in time.
I consulted for a bunch of companies, brought YPC Technologies, a food robotics company through the Real Ventures FounderFuelaccelerator here in Canada, then became a mentor for Techstars, and joined a Y Combinator 3D printing company, AON3D, as the head of engineering for a while until they grew to their series A, so I left them and joined Halodi Robotics. Helping take them through a similar process, growth phase where you double to triple the size of the company in a matter of months, and scaling up the technology to take it from concept to production.
What is the mission of Halodi and what are your specific use cases?
The mission is to bring safe and capable robots to everyone. We want to bring these robots to home and health care in the next 5–10 years. Our dream is to have them as nursing assistants at home, to take care of the elderly. But that is a complicated problem. Home health care is an unstructured environment, which isn’t good for robots. We’re taking a step-wise approach, we want our robots to integrate into the human world instead of changing the human world for robots. When you build a factory, you often design them for the machines more than they’re built for humans. So people are a bit of an afterthought, you optimize it as much as possible to be a black box where raw materials go in, and widgets come out. But this is not a collaborative environment at all.
So our approach is to build a robot that interfaces with humans. From how we interact with environments to how we manipulate objects, etc. That’s why we have a humanoid robot platform. A stepping stone towards that dream of home and healthcare. We’re taking on an initial wave of let’s call them easier applications.
What have you guys accomplished so far?
The company’s been around 5 years, and we’ve worked with DeepMind, NASA’s Jet Propulsion Laboratory (JPL), Toyota Research Institute. Over the past year, we pivoted and targeted the security guarding and retail markets to see if we can:
- Has the robot patrolling hallways at night, i.e. check if doors are closed? Being an intelligent video camera with superhuman capabilities?
- Throw in carbon monoxide sensors, thermal vision, intruder detection, noise detection, etc., but also have the manipulation capability that a simple tower robot wouldn’t have?
- Does this robot stack shelves in a store? As a customer assistant if needed, and then the food packaging and more manufacturing world, so can we have this robot on the floor and production facility, helping with the boxing of objects being technical support or remote technician on machinery and devices?
- We have an avatar mode on our robot that lets users teleport into the robot using off-the-shelf consumer VR tech. Like an Oculus Quest2, you could take full control of all the 24 degrees of freedom with a robot, manipulate objects through the robot in the real world, have bi-directional communication, and understand what the robot observes.
In the manufacturing space, there’s machinery around the world and I can have my expert engineer in one location and perform tech support somewhere else. In the security guarding world, you can have one base station, like one security operation center, and robots patrolling around with operators teleporting into the different robots as required to resolve certain security events.
This lets us do what autonomous vehicles are doing, automating everything up to the last mile. So like Uber, Tesla, etc., can you automate the 80%, the highways, and then humans join for the last 20%? The same feature for us is can we automate 80% of the tasks with the robot, and then when the robot comes across something it hasn’t been programmed to do, a human back at a base station takes over VR style and resolves the event.
Right now, we’ve partnered with ADT Commercial security group, Altopack in Italy, for the food packaging and manufacturing role. And we’ve partnered with StrongPoint in Scandinavian and Baltic countries for intelligent retail and that type of application.
What’s different between you and other Robotics-as-a-Service (RAAS) companies?
We are the only humanoid robot commercially available in the market. Considering Boston Dynamics’ Atlas, which isn’t commercially available yet. We take a very Norwegian/Canadian approach, as a helper first robot. It is in our shareholder agreement that our robot will never be weaponized and/or military applications.
We also create a holistic design for the entire robot, full vertical integration, where everything is our custom design. When engineers started developing the system, they said off-the-shelf motor technology isn’t fit for collaborative robotics. So we offer another product, which isn’t that advertised, our Revo motor, a motor kit for direct drive. It is the highest torque to weight ratio on the market right now. That lets us do 1-for-1 transmission ratios. And this lets you have more sensitivity when doing robotics and let you actually back drive the motor.
Right now, factory robots are geared up to the max with 100-to-1 gear ratios. This lets them have considerable strength in small packaging. But on the downside, there’s no sensitivity if you hit something. Let’s say your encoders, on the back end, can’t back drive 101 gear ratio, especially if they’re all like planetary gears in there. Most collaborative robots in the international market have all sorts of torque and force sensing on both ends of the transmission so that you detect when you hit something. Then you could have a safety watchdog on the control side to stop you from moving. But the microseconds or the milliseconds of delay can cause significant damage and injury, right. A 1-to-1 gear ratio can back drive the robot, you can have low inertia in your movements much like a human. But because we have such high torque, we can have the strength of a human as well.
I know scaling is a substantial barrier to RAAS companies. How does Halodi plan to mitigate this?
So let’s take the dream mission of home-run healthcare, and we don’t have the answer yet. We know that the regular consumer is a lot more price-sensitive than the security company. The applications we’re targeting right now are where the unit economics justifies it. The costs of a security post are expensive. Activities like patrolling, responding to events, or tailgating, etc. These are all key-value tasks that we can put a price on. But people don’t realize a price can be put on them yet. The security guard industry has a nearly 100% churn rate. They’re constantly retraining security guards yearly, monthly, every quarter at times, and so that that cost overhead is large. And then there’s the quality cost, a ramp-up period before your security guard is well trained. With a fleet of robots, you can avoid those costs. You can copy, paste, and redeploy your software application. Across more robots, it becomes more linear unit economics at that point. If you want more guards, add more robots.
Retail margins are razor-thin, so anything for them that gains someone to 2% cost reduction is worth it. They are a high revenue low margin industry. So a few of those percentage points wins it for them. Now we’re noticing the same method applied to retail and other industries, where you have your shelves and all that being replicated.
Humans might be adept at it from working with unstructured environments and situations. But we’re also a lot better at higher-value tasks, like the customer interaction side. Also, robots cover the health side, where they are momentous for health and safety. All that repetitive motion. There are admirable studies about the costs on employees, the cost of employers who have physical strain and injury in those situations, and how robots can bring the health and safety side of it to that.
I’ve read about how a solution to helping robotics scale is wireless power. But simultaneously, I have concerns of the potential radiation from radio frequencies (RF) that would come out of that, so I’m curious to make sure you’re staying within the ethical boundaries?
I don’t see wireless power being in facilities in 3 to 5 years to release in like. We have to depend on available technologies. When the day comes to that, we might pivot. It would be wonderful to pivot towards wireless, the amount of weight savings we can have would be awesome. There’s a reason why the batteries in our system right now are the lowest point on the system. Because robots have to be self-balancing. Any higher in the center of gravity will make our balancing control a lot more difficult to control. More mass means more energy consumption. So, if we could remove all that mass, not only will our energy consumption be reduced, but the robot becomes safer because you have much lower inertia for any collisions that might occur.
The amount of compliance that would be required to bring in any new wireless RF technology into those facilities would be a lot. And what would suck for us is we don’t want to be the drivers of this new crazy technology for a third-party company, unless the contract is enormous of sorts. Some of our pilot partners we are talking to say that to connect to their Wi-Fi is a 6-to-12-month approval process, never mind the distribution of RF.
That’s a tough situation with deep tech, you want to take the full innovation approach. You have to realize requirements have to be based on reality if they ever want to see the light of day. Wireless connectivity is important, where most people come out of R&D with these MVPs. But then they go to an aerospace facility that doesn’t want Wi-Fi.
When I was in additive manufacturing I was at the ISO and ASTMconferences, writing the new standards around health and safety, like 3d printing. Schools were starting to make safer spaces for 3D printers because they fill up rooms with carcinogens. I find it funny with startups when they sprint to build their MVP, bring it to a couple of beta customers, and then hit production only to find out they’re not being compliant. It’s a major roadblock when they realize they have to scrap half their design because none of that is compliant. It’s an incredible task to bring your device through the directives of compliance for North American or Europe. You’re allowed to sell your device in mass quantity to make sure that your device is safe.
That’s why we test. To make sure that we don’t get into this situation where our equipment could interact with other equipment, and cause physical harm. Certification compliance can cost upwards of $100,000 in six months. And that’s, that’s not an agile process at all. People getting off the ground like that drain the bank quickly, especially if you’re pre-seed.
With all those challenges in mind, do you suppose a deep tech company can ever be agile?
I am a firm believer in the whole agile concept of test early, test often. Our hardware development operates like a software sprint, where every two weeks we 3D print the whole robot body. We are performing integration testing continuously, as quickly as we can.
With agile, while you’ll only go into production and certification at a later date, nevertheless, it’s a user story that you need to integrate early on. You shouldn’t design yourself into a corner that you can’t comply with later on. You don’t have to take care of it immediately. My biggest advice on mentoring other startups and founders is don’t go into full scale or presume you are in production too early. I’ve seen many companies that get some traction, and start selling their R&D units as if they’re in production. This brings a world full of hurt when you are a leader supporting hundreds of units out in the field too early as R&D beta units. But because of the smell of money, saying no to sales is perplexing. You have to be vigilant. It’s a short-term game for long-term costs. If you’re waiting for technicians to fix devices across the world, the customer experience, in the long run, will be unsatisfactory. They’ll feel like they’re getting R&D out of the academic lab, so you’re not going to get your word of mouth marketing. You’re not going to get that satisfaction for that Net Promoter Score (NPS) that you need to drive your traction forward at a later date.
Where do you see the RAAS heading this decade?
White papers have come out for what’s expected of Industry 5.0. Where you connect everything with 5G wireless connectivity. Then Industry 6.0 would be almost no code, the concept of no code applied to robotics. Where anybody off the street can spin up robotics without service integrators.
In the past, the rule of thumb was that the cost of the application was 3X the cost of the robot. So a $100,000 robot, at base cost, would require you to pay ~$100,000 for tooling, jigs, fixtures, work, whatever you need to build around the robot application. Then, you pay another $100,000 for the integrator, the software engineer that will program your application. And like that, what you thought was a win at $100,000 became a $300,000 project.
In Industry 4.0, you’ve knocked off the pain of a system integrator to do it for you, because now your technicians can do most programming. But now I’d love to see it knock off all those costs, so anyone can build robots like Legos. Where you buy some sort of pick-and-place software module that you plug into the robot with a USB, and build with off-the-shelf components. Now my robot has navigation capability because it’s all plug-in and play. That’s my real dream. Where you take everything from the factory to your house. Now you bring the democratization of digital transformation to the masses.
With democratization, innovation will accelerate at a faster rate. There’s a lot of capital expenditure (CapEx) and operational expenses (OpEx) in hardware. The amount of cash flow you need to buy your inventory before you even earn money off selling the final product creates a barrier to entry. It’s a serious strain on even full-fledged companies. But in a future like that, all the hardware becomes a commodity. Now you can drive innovation because you’ve also reduced the barrier to entry on the cost side.
We want to bring humans to focusing on higher-value tasks. Humans are much better at application engineering in customer support, those types of situations where that human interaction element, that creativity, that intuition is a key driver, not the rote tasks of pickup, drop-down, pick up, drop down, screw-on assembly. And those are the tasks that cause significant injury, and things like that. There are opportunities for greater advancement in those areas. There’s always the other approach, the Red Hat, the open-source model of companies where here’s the entire tool you want open-source, and you could do it all yourself or pay us, and we’ll manage it for you. Like what IKEA does with furniture, but with robots.
If someone with no formal degree wanted to do what you do today, what would you tell them to do?
I took a roundabout approach, which makes me a generalist with software and hardware. This allows me to lead the team, so I focus on hiring specialists for each of the departments. When I went to school, McGill and ETS had different mindsets. McGill was the old school, Ivy League style. You are special because you are at this university. They had a robotics lab, but no robots! Everything was theoretical. ETS started as a technical institute for tradespeople like electricians and machinists who want to get a more technical degree to become engineers, so they were hands-on from the start. Their lab had more robots than we knew what to do with. Students were programming full-on robots from day one, and you had to complete at least 3 internships to get your engineering degree. So you come out of the school with full work experience.
I love hiring fresh grads from there because I know they’ve not only had a good degree, but they have work experience. Proof of work is so key, especially at a young age, it forms your education. So you can know why you’re learning this material in class. At McGill, I was learning fluid equations and asking myself where I would use these? They also had mechanical engineers learn Fortran and C, telling us those are the 2 most important languages for our future. And I laugh now looking back because some basic Python would have been so much more helpful for everybody’s careers going forward.
On the software side, you definitely don’t need a degree. I don’t ask for degrees when I’m recruiting anymore, because I’d rather see you have experience with open-source technologies. You can pick up ROS, the robot operating system, software stacks online, spinning up in an Ubuntu VM, and you’re ready to roll with robotics. Throw a little Python stack on top of that, and you can be doing reinforcement learning overnight. It’s incredible how quick that is. Then you can buy a Raspberry Pi and camera or Arduino for $30 to work on computer vision stuff. The algorithms are agnostic to the accuracy of your interface. If you’re proficient with the algorithms, I’ll give you a better camera. But it’s the core foundation of how to do things and why to do them.
What are your thoughts on Diversity & Inclusion (D&I) in engineering?
I’ve been doing some talks on recruitment and diversity because engineering sucks at diversity, We’re trying to take our approach on recruiting away from asking questions like:
- Did you use this one specific tool at this one company in this one application?
- Do you have this degree from this one university?
It’s a privilege to have access to all those things, but it’s a barrier to entry for many people. For the past few years, I’ve been trying to pivot all my job descriptions from being on these requirements and focus on can someone achieve x, y, and z. I don’t care what you’ve done before, or what degree you have. But are you the person that can bring me this value over the next six months? Show me GitHub projects, or a blog with photos of a recent project that demonstrates your work.
Nowadays, it’s easy to get your hands dirty with core technologies to demonstrate to startups. These big tech companies startled me because every time I did due diligence I learned the core piece of their infrastructure underneath is merely a Raspberry Pi, or an Arduino.
I don’t expect everyone to be an entrepreneur, as much as I love the mindset. I almost don’t hire an entrepreneur, because too many cooks in the kitchen might stir issues. One of the engineers I recently hired had a Starbucks Master Brewer certificate because he wanted to do that as an extra activity. And that’s a mindset I wasn’t keying in on for all my other interviews that he took to show nontraditional hustle.
When younger people ask me what degree they should go into, I like to flip the question and ask:
- How should you apply to this company?
- Who do you want to work for?
- Who works for that person?
- What are their values?
And then I tell them to work backward from there. Don’t assume, with an electrical or mechanical engineering degree, you will automatically fit in Google. You have to work backward. It’s almost as if solving for x is a lot better than inputting x regarding your career.