Episode 98 – Autonomous Cleaning Revolution With Austin Deric

Commercial Robotics Automation: In this episode of Over The Air, Bill Brock, CEO of Peridio sat down with Austin Deric, Director of Software at Tailos, for a fascinating journey through the evolution of commercial robotics. Austin shared his path from the innovative halls of Southwest Research Institute to leading software development for enterprise-grade autonomous vacuum cleaners that are transforming how commercial spaces approach cleaning operations.

The conversation dives deep into the practical challenges of building reliable robots for real-world environments. Austin reveals how Tailos’s approach differs from consumer robotics, emphasizing their philosophy of “out-of-the-box autonomy”. We explore the technical architecture behind their success, from their transition away from Raspberry Pi constraints to embracing open-source solutions like ROS and SLAM Toolbox.

Austin offers valuable insights into:

• The critical differences between contract engineering and product development.
• The technical evolution from legacy systems to modern robotics architecture.
• Real-world challenges of enterprise integration and scaling.
• The thriving robotics ecosystem emerging in Texas, particularly in San Antonio and Austin.

This episode offers valuable insights into the real-world application of cutting-edge robotics technology. Austin’s combination of technical depth and practical business acumen makes this a must-listen for anyone interested in the commercial robotics automation world.

About the Show

“Over the Air” is a podcast that features unfiltered conversations with executives about their IoT journeys. The mistakes they made, the lessons they learned, and what they wished they’d known when they started. In each episode of Over the Air, our expert guests make bold predictions, reveal unpopular opinions, and share stories you won’t hear anywhere else.

Buyers Guide – How to Choose the Right IoT Company: Very

Subscribe here

YouTube
Apple Podcasts 
Spotify 

Connect with Austin Deric

Tailos  Website
Austin on  LinkedIn

Transcript

Hey, Austin, welcome to the Over the Air show. Hey, Bill.

Thanks for having me. Very excited to be on. Yeah, and for those listening today, we have a really exciting participant, Austin Derek from Talos.

Talos is a commercial-grade robot vacuum cleaner, and Austin heads up the software team over there. So a lot of experience and lessons learned in the robotics and distributed software space. So really excited to get into it today and learn a little bit more about you and the company and taking this product to market.

Yeah, I love yapping about robots and product and all that good stuff. So I’m very excited where this conversation is going to lead. Cool.

Well, before we get too much into vacuums, would you mind just telling me a little bit about your background, where you come from? What gets you excited? Sure, sure. So I’m director of software at Talos. I started my journey actually at Southwest Research Institute here in San Antonio, working very closely with the robotic software developers there in, let’s say, 2008-ish, that time range.

So if you think about what the world of robotics was like back then, Ross was, I don’t even think it was, it was like pre-release. Willa and so that technology stack really solved a lot of our pain points that we were having when we were kind of doing these robotic projects and inventing a lot of stuff from scratch every time. So this kind of unified framework for robots that we could all open source develop together as a community was really formative for me.

So like I just entered my career and this thing was just coming into the stage. So I really leaned hard. I loved it.

I was sold. So I really leaned hard into that and essentially did robotics on a project basis up till about 2016, working in their industrial robotics group, Ross Industrial, massive shout out to them. Just working with amazingly talented people.

So was this, so in Southwest Research, they were a contributor to Ross or what was that like? Anybody can contribute to Ross. We did contribute heavily. There’s also SWEERI.

We call it SWEERI for short, Southwest Research Institute. But they also maintain their own packages and offerings around industrial robotics. They also have an autonomous vehicle too as well.

So they provided their own packages and contributed to the actual Ross itself. And the Ross Industrial offered a very tailored solution for industrial robots. So like serial manipulators, things like that.

And really, one of the important things about open sources, communities like that is really just onboarding companies who we kind of do it differently and have their own legacy technology. So really giving, putting Ross applications in their hands and then they, so you kind of increase usage and then the ecosystem churns and great things happen. And SWEERI, they’re a .org? Yeah, so like a 501c3, yeah, nonprofit organization, like project-based, contract-based engineering services, essentially.

So customers would come, hey, I really need a robot to clean an airplane, take the paint off of an airplane. So we would kind of create solutions for them, the robot, the software and things like that. Yeah, that’s really cool.

I was the first I’ve heard of them. That’s very interesting. Yeah, it’s a really amazing underrated place with a lot of talent.

The group of people during my time there I worked with have gone on to be CEOs of companies, startup founders, making incredible amounts of money and progress in itself. So kind of that network scene, what they’re all doing right now is quite amazing. But we were kind of so successful and it was so great.

And we were all doing a great job, but the problem with contract engineering is it doesn’t scale. Essentially, it scales with your time. If you want to make more money, you got to hire more people and do more work.

I’m familiar with this problem. Yeah, so we really wanted to do products, right? So we have this great technology. You can make a laundry list of product ideas and you start picking off the low hanging fruit, essentially.

And so a lot of us were kind of talking about this and we essentially went off and did that in our different ways. Plus One Robotics was founded, which is essentially a kind of logistics pick and place style product offering. Extremely successful.

So I joined them for about a year, developing some of their kind of human in the loop pick and place AI products. Had a great time, working with a lot of the same people on very… Actually, their problem sets actually compared to the one now is actually really easy, like nicely controlled, great product market fit. That really taught me a lot about actually the product development cycle and how to actually make a product.

Because to date, we’re only on project basis, not product. And so it’s wildly different. Taught me to obsess a lot about the customer and deliver product fast.

And what was your customers there? You mentioned pick and place, but that’s kind of a wide use case. What were you working on there? Yeah, so essentially they’re logistic companies and kind of large retailers who have big logistics operations. So they have a lot of packages going everywhere and they need… Before Plus One, people grab in packages, putting here and there essentially in these big warehouses.

So kind of coming to market with a AI-based human in the loop product offering that could do throughput faster than humans. They don’t take breaks. You don’t have to scramble for… When you get to your peak seasons and things like that, you just turn the robots on and they work at higher liabilities and faster throughput.

So the product market fit was incredible, hugely successful. And it taught me a huge amount. So I kind of took those skills to Talos and essentially rinse and repeat.

At this point, I kind of specialize in startups and robotic startups. So essentially take those skills, build the team, build the product, have it be commercially very successful. So what’s the secret sauce around Talos? Or what is so exciting about this application? Great question.

So essentially it’s… I have one here. I love the product. I fell in love with the product.

They were a raw space company. So I came in a couple years after they started. But essentially this is an autonomous car.

You get your lighter, your camera with just a vacuum attached. So that’s really fun problems to work on, essentially like autonomous navigation in indoor spaces. The challenge, the problem they’re solving, unlike the previous ones with industrial robotics, it’s just very difficult.

The environments you’re operating in are wildly different and they need to work to the customer’s expectations and chaotic environments, different environments that we’ve never seen before. So yeah, I’m hooked. There’s no shortages of things to work on.

And the form factor is cool. What’s this spinning piece on top? Yeah, so we have an RP Lidar. This is an A series from them.

This is kind of one of the older products that I just have for internal testing. But we have kind of… The Hipper Podcasts. Yeah, yeah.

Of course, the industry standard real sense. It’s not the best depth camera in the world, but it’s kind of like the gold standard. And I’ve used them kind of reliably.

It’s kind of like IBM. You never get fired for having an IBM, but it’s super reliable. I think that in terms of making products, one thing that’s underrated is just reliability.

So sacrificing performance for just a product that you know that’s going to work every time. We have enough reliability problems to solve that just not having those is huge. So what about this product is so reliable? So you guys are mostly… I’m looking at the videos real right now.

It’s like hotels, airports, things like that. So what are some of the features that you’ve built in to make this work well in these commercial environments? Yeah. So we have a few philosophies that are kind of guiding product development.

The main one and the one that our customers really like is the ability to take the robot out of the box, put it in any location, put a battery in, press the play button, and then essentially they can kind of walk away and go do other things for as long as the battery lasts, which is an hour and a half to two hours. Then the robot cleans the whole area, they get a notification, okay, come change the battery, they bring the battery, they change it, resume the run, clean the area. There’s no setup time.

And then they can go view data through the maps and the dashboards and understand kind of how looking back on their day, how the robot cleaned and then kind of where they should clean next and things like that. What’s been different about this? I mean, everybody’s kind of look at this and compare it to obviously iRobot and then the other like short and interproducts in this space. What’s different about the commercial application versus consumer, would you say? You know, I don’t generally, I don’t think a lot about the competition and kind of non-use cases.

I don’t know if we would ever make a consumer robot, kind of really focusing the company on kind of B2B like enterprise sales, that type of stuff, those customers. And also given their scale and kind of working closely and integrating really like enterprise integrations. So kind of to not answer your question, I really don’t know.

I don’t follow iRobot a lot. I’ve used their product. It’s great, but just different use cases, different customer sets.

So why I’ll be different. So what are some of those integrations that you guys have had to perform on the enterprise level? Yeah. So a lot of our customers, so our customers actually similar to the environments we operate in kind of run the gamut of sophistication.

So the more sophisticated customers will have complete IT teams and complete back end systems and databases managing their operations. And part of their operations are cleaning, right? Cleaning guest rooms, cleaning hallways. So they really demand this data so we can tell what they’ve cleaned, what they haven’t cleaned, the usage, the cleaning people are operating, how they’re performing.

So we essentially have a huge amount of data we collect off the robot and we can take that data and share it back with our customers and drive value. Have there been any leaps or developments since you’ve been there that have been really exciting or transformative to the products or customers that you can service? Yeah, I would definitely say so. You know, we have an amazing robotics development team.

So when they kind of make some changes, it can be extremely impactful. So kind of when I came into the company a couple years in, we actually had a completely different version of the product. It was a little bit smaller form factor.

It couldn’t do these like large open spaces, which you need to, you know, more battery, bigger, faster, to be able to clean it in reasonable amounts of time. So when I came in, we had this, I’ll call it like the legacy code base. You know, I have a version of that product right here.

It has essentially, same letter, no camera, essentially it has a Raspberry Pi 3 in it. So the code base was built around that. So kind of imagine you’re, you know, you’re gathering your lighter data points, which are massive, you know, your particle filters, and just the localization using an RP, you know, Raspberry Pi 3 is just really challenging.

But our code base was really tuned to that. So bringing on the camera, increasing the actual CPU performance, we use a Toradex iM8 song. It’s like an ARM processor with like six cores, you know, four cores are low power, two cores are high power, eight gigabytes of memory, which Raspberry Pi had one, you know, so that makes you make a lot of compromises on the code base.

Yeah. So upgrading by making those hardware upgrades and adding the camera, and we could really be more flexible in our code base and do things like we use slam toolbox, essentially, you know, as our local. What is that for the yeah, for those who don’t know? Yeah, so essentially slam toolbox was developed kind of through the Ross ecosystem.

They actually, you know, they were so successful, they kind of branched out and made their own startup. I’m forgetting the name I’m bringing out all of a sudden, but Steve McKenzie, I believe is leading that. So really happy for his success, too.

And his success is our success, because, you know, we use his software, we kind of grow, we contribute back. But essentially, it’s a post graph localization system. So this is how the lighter, you know, gathers points, and then kind of creates the localization maps that it uses to know where localization, you know, basically, it’s like, where’s the robot, you know, in this hallway or guest room.

So, but these post graphs, and these computations take a huge amount of resources. But you know, with the bigger CPU, now we can do things like that. And we can kind of have much more logic in our kind of our planners, so we can, you know, create very beautiful cleaning paths.

Nice. When you’re looking at tools like this, and maybe other autonomous applications, how do you guys evaluate build versus buy? Because I know that like, there’s, I talked to a company recently that’s doing something, I would say similar only, and that it’s like an autonomous something, but they’re doing airports. And I think that they were a little hesitant, not cleaning them, but operating outside of them.

And they’re hesitant to pull in something like SLAM, I believe. So how do you guys think about what you’re willing to bring in the efficiency gains, etc? Yeah, that it’s an excellent question. So kind of the legacy code base was an understandably reason, you know, for the pie, it was very custom, it was it was huge and complex.

We had the we had our own localization package that, you know, use particle filters, we called it Elma. So we pretty much inherited a 100% make decision, right? But, you know, we really wanted to do a lot more development much more quickly. So it’s essentially, if you can’t, it’s a buy decision, unless like, you just can’t make the buy work.

And the great thing about open source is, you know, you can actually like, do the buy decision if it doesn’t exactly meet your use cases, which it didn’t, and we had to make some upgrades to it. But essentially, we did that. And so it was kind of minimal.

And we actually contributed upstream. So now we don’t have to maintain it at all. So over time, most of our stack is kind of a buy decision, really heavily influenced by nav two, and what they have going on in their sort of offering, you know, product off on the data side, and, you know, you can make your own EC2s and databases and, you know, dashboards and things like that to view your telemetry and your testing.

Essentially, we’ve essentially it’s like an all by decision on that to kind of using, you know, software service approaches like timescale database, we have Grafana dashboards. You, you’re really plugged into this hardware ecosystem. And I loved hearing about the history and what I’m hearing a lot of companies coming out of Texas in the robotic space.

And am I just biased because I’m working in hardware and trying to talk to folks like you? Or what do you, what do you see going on around you in like San Antonio and Austin areas? It’s a hotbed, you know, San Antonio will give Austin a run for its money in terms of like robotic talent, and startups and things like that largely due to Southwest Research Institute, just like spinning off all of this great talent and product ideas. And, you know, once you get critical mass, you can kind of start building a community around it in the cities. You know, you got, you know, plus one, obviously, in San Antonio.

Actually, Dark Hive is a new kind of UAV type company. Just moved to San Antonio. Really excited about what they’re doing.

You know, in Austin, of course, you have Talos Robotics, Fox Robotics, Aptronic, you know, the humanoid robot. Yeah, yeah, yeah. There’s a lot of very impressive work happening right now.

It’s crazy. Yeah, I’m very excited. I think a lot about them.

That’s a, that’s, now that’s a hard problem that you, you know, I’m not the biggest fan of humanoid robots, but I see all this activity and funding and talent going to it. I kind of question my assumptions around it, but I wish them the most success. And, you know, I know they’re using the kind of Ross tools and contributing backs.

You have Aptronic in Austin, Agility Robotics in Boston, somewhere in the Northeast. Melanie Weiss was actually a Willow garage, I’d say founder, actually a really big inspiration for my career. I’ve had the opportunity to work with her.

She’s kind of advised us in many ways. So she’s the CTO of Agility. And actually, one of their competitors, Aptronic in Austin, actually, they had, they got a huge new talent when one of my old mentors from Plus One, from Sweery, is now their Paul Voss, head of their product team.

So you can kind of see how it’s such a small world, especially music and these only talented people. And yeah, just really excited about the future and what they can do. What industries do you view as the best opportunity for robotics over the next five years? I mean, humanoid is really cool.

Obviously, if we can replicate people’s capabilities, that is what everybody’s going for. But before that, what are the big growth areas that you’re seeing? Yeah. I’m still not super clear what problem the humanoid companies are solving.

I think that to get an MVP, or whatever they decide their problem is, is at least 10 years away. Just the grasping problem alone is such a hard problem. That’s not solved.

But generally, when you want to bring robotics companies to market, you really want to take the easy problems to solve problems. You don’t want to take on the grad school project. The grasping problem has been a grad school project for like 20 years.

So what are the easy unsolved that you’re excited to see tackled? I think autonomous cars are the obvious one. I still think vacuuming is, there’s a lot of room to go there. You should go into any space and see robots cleaning the areas.

There’s such labor shortages around to actually work with people to increase cleanliness, especially around COVID. And if that could happen again, essentially, what you’re cleaning robots are is like endpoint sensors to collect data. So maybe they can collect viral data and an early warning system essentially for public health and things like that.

Your last mile as well, autonomous car and that last mile problem. I feel like it could be solved. And there’s a lot of companies working on it.

Amazon’s got a few offerings. So yeah, I think those are the big ones. But honestly, I’m so focused on getting our products developing on features.

And I really just like laser focused on that. So don’t really kind of think about the other things too much. Well, you guys clearly have your plate full.

So personal issue with these robots. I had a, an I robot and I had a German shepherd at the same time. And those two things were not compatible.

And so it just wouldn’t kill, it would just fill up so fast because those dogs shed so much. How do you guys handle, you know, you know, how do you really optimize the environment for the capabilities of the actual vacuum? Yeah, that’s a great question. So essentially for things like that one specifically, so we have replaceable bags that are larger in volume for commercial for our commercial clients, but also kind of just integrating with their operations and like the amount of cleaning we do in any given area, like this, my consumer is a little bit different, right? You might run your vacuum, you know, in your house once a week, like, and that would probably be a lot for most people.

I’m actually, it’d be interesting to know what, you know, cleaning frequencies are amongst consumers, but you know, our, our commercial clients, you know, they have staffs that are constantly cleaning. They’re just running. Yeah.

So are really what you get to like the first, you know, when you first go to a property, it’s a completely different story because you just, you wouldn’t believe how much stuff you collect and our customers are actually like shocked. They’re like, Oh my goodness. Like I had no idea.

But over time, once you get through that transient over time steady state, it’s really, you just, you’re keeping a very clean hallway. So you don’t have as many bag fillings as you would imagine, like kind of once per day, like just, you know, new bag at the beginning of the day. And that’s enough to go through.

But I’ve seen a lot of people with docking stations also like, you know, go back and dock and empty out the, I think I robot actually has that too, the empty out your, your, your debris. Or so they would empty the actual dispenser. That’s cool.

Yeah. Yeah. I’m looking at this product.

It looks, it looks very robust. And I think that’s like some of the difference between the consumer versus commercial. And I mean, you guys are building things that need to just run 24 seven.

And that is, that is very clear by the, by the form factor and kind of the, the sturdiness that is just really obvious here. Yeah. Yeah.

I think we’ve had it in customer hands for about two years now. We have a lot of customers, those original customers running it, you know, really a lot in those early robots are still, they’re chugging along. Nice.

Really happy. How long did it take? How long did it take them to get from early product to some sort of, you know, market market reach to traction? Yeah, that’s it. So we, when COVID hit are really transformed our business.

So a lot of our customers are hospitality customers. And so COVID pretty much shut down their businesses, right? So that actually gave us time to, we knew we wanted to make a different product to kind of essentially, if you look at like a market share within a hotel or an airport, like we wanted to make a robot to go into more places and clean more carpets of more people. So we essentially started from scratch in that.

And it took us about, I would say, what, 2020 to 2022. I remember it was really zero dot 13. And let’s say October, 2022, where we first deployed it with a customer, a paying customer who really enjoyed it and gave us great feedback.

So I kind of call that like the MVP. So essentially, you know, COVID started to October is when we kind of hit the MVP inflection point, and then kind of started scaling at that point. Nice, nice.

Well, it’s a beautiful product. You guys are clearly, clearly doing great work. Yeah, amazing, amazing team.

Yeah, I love working with, with, you know, our developers, they’re just so incredibly smart. And they’re passionate. And that really feeds me.

And they teach me a lot. That’s, that’s the, you know, the most exciting thing is that, you know, they’ll evaluate a new technology, or hey, we can do it differently like this. And it just makes every day just really kind of awesome.

And you never know, you never know what you’re gonna get me, you know, at the end of the day. Awesome. Well, I think that’s a really nice note to end on.

I really appreciate you coming on Austin, really, really amazing story and great product and would love to stay in touch and have you back on in a couple years once you’ve built out the humanoid version of this. No, I’m kidding. But next, next product lines, and other, you know, other things you have on the radar.

So we’ll keep in touch. And also I need to make it down to San Antonio and see what all the hype’s about. Oh, we’d love to have you.

Great. All right. Well, thanks a lot.

And we’ll, we’ll pick it up soon. See you later.

LEARN MORE: Discover the essentials of IoT connectivity with our comprehensive guide—an invaluable resource for anyone looking to navigate the complexities of the IoT landscape.