In a world where disc-shaped robot vacuums still stumble through homes, blindly bumping into furniture and struggling to not get tangled up by wires. Matic Robots is taking a fundamentally different approach. Founded by former Google Nest engineers Navneet Dalal and Mehul Nariyawala, Matic has spent seven years developing a vision-only system that lets robots truly see and understand our homes.

In our first AMA session (you can check the full video here), Mehul shares insights into why current home robots have failed to meet consumer expectations, and how Matic's human-inspired approach is changing the game. He also reflects on his own entrepreneurial journey, sharing lessons learned and advice for those looking to build something new.

The following insights are adapted from a recent live Ask-Me-Anything session dated on 03/27/2025 with Matic co-founder Mehul Nariyawala, where he answered questions from robotic enthusiasts and Matic customers. This blog has been edited for clarity.

Rethinking Home Robotics from the Ground Up

Q1: "What's the most surprising obstacle you've encountered in making robots truly autonomous in unpredictable home environments?"

Mehul: Before we left Nest, we used a bunch of robot vacuums and realized—you know—Dyson has amazing suction power and battery and vacuum components. They’re a phenomenal mechanical design and mechanical company. But when it came to the algorithm, that vacuum was probably the worst robot vacuum I’ve used. Simply because nine out of ten times, it wouldn’t even be able to find its own dock. So that’s when we realized the computer vision approach Dyson was taking was potentially the right one, but they were really lacking. The second thing was it got onto one of our really nice rugs. And because of the suction, it got stuck on it. We weren’t at home, and it kept the vacuum on for 45 minutes. When I came back and picked it up, I realized the entire patch of the rug was gone. That’s when I realized—actually, it didn’t even know it was stuck on the rug. It didn’t have any understanding and just kept going. It really ruined the rug itself. And second, maybe the suction Dyson talks about isn’t needed.

So those were some things in the back of our minds. It piqued our curiosity—like, why isn’t anyone tackling these home chores, home robotics problems? Why aren’t there robots there? So Navneet and I spent about two months just digging into robotics itself. Prior to that, we had no background in robotics. We just spent a couple of months doing it and very quickly came to this conclusion—this was our conclusion—that the entire field of indoor robotics, specifically robots in homes, is built upside down. It’s like putting the cart before the horse. The reason I say that is, even though self-driving cars require all these sensors and vision and amazing algorithms, they have two things: Google Maps and GPS. At least they know where they’re located on the road, which road, and where it’s going. On the flip side, when it came to indoor robots, the analogy we realized is that back in 2002, Roomba did a phenomenal job when there was no AI or technology available. They built a robot with a blindfold on that would just bounce around and clean our home. That was a great way to start robotics. But then along the way, more sensors got added, but those sensors weren’t really converging into intelligence.

The robot didn’t understand which part of the home it was in—whether it was the kitchen or dining area, right outside the living room, or the right side of the couch or the left side. So we realized, look, if you’re gonna build fully autonomous indoor robots, we should give them the same perception system we humans have, which is two RGB cameras and a bunch of algorithms. This is because—and this is the insight Navneet and I have been operating on—the indoor world, specifically homes, is built by humans for humans to fit our perception system. It’s optimized for a vision-based system. So if we’re gonna have Level 5 robots—and the definition of Level 5 being that if Level 5 cars drive like humans, then Level 5 robots must behave, navigate, manipulate objects, and do tasks on our behalf like humans—then they need the same perception system as us.

That’s where the decision came from—hey, we’re gonna do the whole thing on the device using cameras only and see if we can solve the problem by giving them a similar perception system. But along the way, we realized that if you’re going inside homes just with cameras, privacy is an issue. Latency is an issue because you don’t want a robot that says, “Oops, I fell down the stairs,” or “I fell on your pet or a kid,” because Wi-Fi is weak. And then the third thing we realized is the indoor space is actually far more dynamic than the outdoor space. You can’t just have a static map like Google Maps. You need dynamic maps that can map themselves, so your robot maps itself in an environment. That’s where we decided we’re gonna do the whole thing on the device so the robot can always update itself, observe changes in the environment—the delta in the environment—self-correct, preserve privacy and be really low latency. That was the thought process. That’s how we started thinking about it.

What was the obstacle? Well, everything about robotics. I like to say that if hardware is hard—you know, there’s a cliché in Silicon Valley that hardware is 10x harder than software—I’d say robotics is probably 10x harder than hardware itself. It’s a much harder thing because we had to do everything from mechanical design to platform to OS-level optimization all the way to algorithms. Just putting everything together so the robot works and behaves in an expected way is quite a challenge in itself. And then the second part was certain assumptions we made along the way.

One assumption was that SLAM—Simultaneous Localization and Mapping, a precise localization algorithm—was theoretically solved in the mid-eighties. There were a lot of papers on it, and the assumption was SLAM is a solved problem. Tons of open-source libraries were available too. So we started thinking we’d be able to use them. But when we dug into it, we realized none of the open-source algorithms had the required accuracy. The best analogy is, you know, touch interfaces were available before the iPhone, before smartphones. But back then, when we used a stylus, those touch interfaces weren’t great. Or if you’ve used an ATM or those early TVs behind the seat in airlines, you had to jab your finger with all your power to make them work—until the iPhone came out and made them amazing. Similarly, SLAM systems were about 70 to 80% accurate, and we decided in the fall of 2020 that we’re gonna rewrite everything from scratch. That was one of the obstacles we didn’t anticipate having to solve, but we did.

From Prototype to Product: Navigating Hardware Challenges with Vision-First Approach

Q2: "Hardware production is a costly, evolving nightmare. How'd you scale from prototype to product–and why pivot away from software background?

Mehul: Great question, Dan. Between Flutter and Nest, we actually went to Nest just because we wanted to understand hardware and we wanted to learn how to do hardware. This came from our time with Flutter, where we did gesture detection over a webcam. It was essentially like Microsoft Kinect, but just using desk cameras that we had. We built these gestures—play, pause, previous song, next song, thumbs up to like a song, shush to mute—so you could control your iTunes and Spotify using those gestures. It was great, and we built this algorithm that became the #1 app in 73 different countries worldwide. But we realized we didn’t have access to the camera. We didn’t have frame rates per second. We didn’t have auto exposure. So it was like building algorithms for an eye that was out of our control and might even waver.

We felt like we’d built half the solution. If we’d had access to the cameras, we could have done a much better job. But at the time with Flutter, we had no background in hardware whatsoever. So one big lesson we took away was that if you’re trying to solve a problem, you can’t arbitrarily limit yourself to hardware or software. You have to ask, what’s the best way to solve the problem? And if that means hardware, then we should go do it. So with the problem we had—this is my own personal problem—I have three kids and a golden retriever dog, and I was using all these robot vacuums. I just realized they were all sort of sensors-first or mechanical-first. We concluded that it required a vision-first approach to truly solve them.

And that meant doing the hardware and solving the problem using a product. That’s what we decided to do. And the reason to do a product as well—this has been, in our philosophy that if you’re trying to genuinely push the algorithm to the next level, then you have to productize it. The only way to know if your research or algorithms work is to put them to the test in a real environment. That forces you to go through challenges you may not have thought through when you’re just doing research. When you publish an algorithm, you’re taking it from zero to 80, but really shipping it to customers requires you to take it to 99.99%. And that means productization. That’s what excites us—getting it into the hands of the users.

What Customers Really Want

Q3: "For consumer robotics, what is an assumption about the customer that was proven false? What was one that was proven true?"

Mehul: Great question! When we started Matic, with a desire to make a robotic vacuum, we had heard from a few folks who had worked at Neato or other places. They had given these long interviews in Wirecutter, and we had read those interviews. They mentioned that actually, navigation inside homes in an unstructured environment is arguably even harder than self-driving cars. And we obviously didn’t believe that—self-driving car mistakes are much more fatal versus inside homes. If a robot made a mistake, it’s not necessarily going to result in any sort of fatality. So we thought that mistakes—there are mistakes like going over dog poop or falling down the stairs, which are really, really bad. We call them “deadly sins” inside Matic. But beyond those mistakes, some mistakes are trivial. So we thought it would be easier, relatively speaking, to get the robot out and improve algorithmically in this space.

And what we realized is that actually, when it comes to robots—and specifically this particular problem of floor cleaning or any home chores for that matter—we as humans and customers, we wanna just delegate, set, and forget. We don’t want to constantly collaborate with it or rescue it or help it. And that’s the part that we—and I think many consumer robotics companies—potentially underestimate.

With AI, if you’re doing coding or if you’re doing any sort of image generation, we tend to collaborate with current AI. We tend to ask a prompt, tell it to make different stuff, and it’s sort of back and forth. And that’s fine. Collaboration with AI is great. It gets us to the task, and we understand that coding is hard. So even if it gets 80% of the answer right, we’re mesmerized. But on the flip side, with robotics, we don’t go to school to learn how to navigate a home without bumping or how to vacuum. A five-year-old or seven-year-old can do it in a pretty decent manner. So in those scenarios, in trivial tasks, when a robot makes a mistake, customers get frustrated quite a bit and think it’s stupid. So there is that balance we have to drive. Accuracy and five nines is really, really critical, and that was one of the things we learned along the way, which we weren’t anticipating.

Q4: " I was skeptical of vision-only navigation, especially in darkness. Now convinced. Can you discuss challenges of on-device AI navigation?"

Mehul: To be entirely honest, that was one of the risks when we started out. We really didn’t know if we could just solve it using computer vision. We felt that was the right approach, and that was the approach Tesla was taking with self-driving cars and full self-driving as well. So we thought there was plausibility in it. But we decided to do it because of a few reasons.

One is that when we looked at robotics and robotics attempts over the last twelve years prior to 2017, we noticed there were Kickstarter campaigns after Kickstarter campaigns that were funded by various robotics companies. But typically, what you observe is that by the time they were ready to ship, either the robots became prohibitively expensive or they never shipped. And this is because each sensor—we had this rule of thumb at Nest, and the Nest hardware experience came in really handy—but we had this rule of thumb at Nest that for a single sensor you add in hardware, assume two or three software engineers on the flip side as a permanent cost. So the more sensors you have, the bigger the team. The more sensors you have, the more complex the manufacturing supply chain. The more sensors you have, the more calibration, the more failure points, the more complex the software stack itself.

So with each sensor you add, complexity rises exponentially. When you take a sensors-first approach, it’s very easy to build prototypes. But then when it came to scaling, we realized that it actually gets much harder. On the flip side, doing a vision-only approach meant that we just have five cameras, and that’s really it. Those are the sensors we have. We don’t even use an IMU at the moment. So a vision-based approach was far more challenging in the sense that initially we’d have to absorb all the complexity in software. And that would take a long time, but if we could pull it off, over time, complexity asymptotes, and software-based updates are much easier than the hardware sensor-based approach. That’s where we can ship a lot of over-the-air updates and continuously improve the product along the way. But also, it keeps it affordable. So that was the reason why we chose vision-only AI navigation.

And it has been hard. It has taken us seven years—precisely that long—because no one has ever done this before at the level that we have, and as optimized as we have, with just four GB of memory on our Nvidia Orin Jetson and Nano platform. It seems like there’s a lot of work that has gone into the software, but I believe there’s been a lot of work as well that’s gone into the hardware.

The Roadmap To Rosie the Robot

Q5: "Hi, are you considering adding an arm to your device? A common use case to me would be to move a moveable object. That's what we do as humans."

Mehul: Yeah, that’s a great question, Nitesh. There are two aspects to it. So one is that I don’t know if an arm is necessary to move—meaning that the robot itself is pretty dense. It has this cleaning kind. Maybe there is a way for us to actually move things out of the way or move things into a proper place even with Matic itself without doing it. And hopefully, you can sort of imagine where our roadmap is going in that direction. So that’s the first part of it.

And then the second part is, there will be a future where we add hands because we didn’t start Matic with this idea that we’re just gonna reinvent the robot vacuum. That’s not what we wanted to do.

We wanted to reimagine, and reinvent it, but that was always step one. All of us who are robotics enthusiasts—we’ve always dreamed of Rosie the Robot from The Jetsons. And that is the goal: can we not have a Rosie the Robot? Can we not get rid of mundane and repetitive chores in our lives? So that’s what we wanna do, and our goal and our approach has always been that we will do it with a product-first approach. So what product can we ship along the way? But also do it the way a human child grows. And this is where, you know, Navneet's computer vision philosophy has always been that we don’t wanna bet against nature.

Nature chose to give us two RGB cameras versus some version of a depth sensor. So let’s go with RGB cameras. Similarly, we don’t come out as a baby and are amazing from a dexterity or manipulation point of view. In fact, what we do is, the first five years of our life, kids just learn how to perceive their homes and navigate from point A to B. And that’s precisely what we’ve solved in the context of the robotic vacuum and Matic. But then as they grow, between five to ten, we don’t give them sharp objects yet. We don’t give them complex manipulation tasks. Instead, what we do is we give them soft, unbreakable items like shoes, socks, and backpacks, and teach them to move those things along the way. And as their motor skills improve, as their perception of the world gets much more accurate, around 14 or 15 is when they become a full human—like a humanoid body form—and that’s when we trust them to do far more complex tasks.

In the same exact way, we anticipate this robotics to evolve in multiple steps. The first step is solving perception, next is solving manipulation, and then eventually putting everything together to say, we have a Rosie the Robot in our home?

Q6: " Of all the exciting features labeled under Coming Soon on the Matic website which one is the main focus of the team right now?"

Mehul: I know a lot of you have been eagerly waiting for toe kicks—which is the space under the kitchen cabinets—as well as improved edge cleaning. For the last month and a half or two months, I wanna say, that’s the only focus we’ve had internally as a team. And we are about to ship a version, hopefully, today (03/27/2025) or maybe tomorrow, as we kinda put the finishing touches on it—which is version, I wanna say, 0.5 of Toe Kicks. I’m super excited for everyone to experience it and push it out there. So that’s number one.

Then from here onwards, there are two or three features that I’m super excited about. One is really this concept of dirt detection and mess detection. And we wanna do that visually as well. This feature was really critical—and is really critical—because we also asked this question a long time ago: what does it mean to be a Level 5 robot inside the home? And as I had mentioned, a Level 5 robot means that it should behave like humans. So in this case, it should floor clean like humans. If it is floor cleaning like humans, then why are our robots sitting on the dock twenty-three hours a day? Or why do they only take off when we give it a command or when we schedule it? Shouldn’t they just, you know, throughout the day constantly explore our home, scout for dirt and stains? And if it finds it, it should just clean it—because we as homeowners and users, wanna live in a perpetually clean home with perpetually clean floors.

And dirt detection is a key piece of the puzzle. Hey, can we just go and look for dirt, find it, and keep it clean? Or if you tell the robot to, say, spot clean, can it actually go into that area that you just pointed out, look for dirt, and keep cleaning it until the stain is gone or until the dirt is covered? So dirt detection is a really exciting part.

And then the second thing is just a very, very easy interaction point of view. Having gestures as a background, and having done human-computer interaction work in the past, we know that natural language instructions and gestures are the way to go. That is the future—that all computing should understand our interaction method. So the first step was to add information like room labels and some other semantics like wires. But now we want to enable very simple commands like, “Matic, go clean the kitchen,” or “Matic, go clean the mess inside the kitchen.” So these are the things that we are really excited about as we move forward.

And then the third thing is, as I mentioned, we believe that there is an opportunity to do manipulation and some tidying up even with the current version of the Matic. So that’s the third one that we haven’t publicly talked about—even on our website—that we are super excited about working on.

Software For The Win

Q7: "How do you balance adding advanced features with keeping home robots affordable and accessible?"

Mehul: That’s a great question, Ken. And the answer is very simply by improving the software. This is why we rebuilt everything on the device.

You know, your computers, your phones get OS updates, and they get better and smarter every year without you actually having to purchase new hardware. So in the same exact way, we wanted to enable—build this vision-based autonomy where software is the key piece of the puzzle—and keep improving it. So our goal is to keep improving the software, but also as we scale, reduce the cost of the hardware, reduce the cost of the sensors. And by keeping it simple—the designs and motor selection and everything along the way—we can make it affordable and we can keep improving it. As we move forward, I believe whether it’s robotic hands or any sort of next robot, initial versions might probably be expensive. But that’s where we hope to take a little bit of a Tesla approach—where the initial, you know, Model S was expensive, but the goal is to democratize robotics and get inside homes by creating a Model 3 or even more affordable versions.

In the same exact way, we anticipate that we will start in a premium category, but as we scale, we wanna keep making it more and more affordable across the board. And that actually applies to our bags and consumables as well. There was a question about whether we are doing a razor—we think of it as making bags or consumables really high-priced is a sort of perverse incentive, in the sense that we want to make sure that you clean as much as possible. So as we scale, we wanna make our bags and consumables even cheaper. And that’s the goal.

That’s the way we are doing it. Even now, we believe that our bags are much cheaper than what might be available out there, with a far better value proposition—including HEPA filters and some of the other advances that we have made.

How is the Vision-Only Autonomy Stack Different?

Q8: " How does Matic's approach to mess detection and surface recognition set it apart from competitors (such as Roomba or Dyson)?

Mehul: At a very high level, we just came to this conclusion that if Matic can see, if robots can see, they can navigate it. If they can see, they can avoid any obstacles. If they can see the dirt versus the mess, they can do it. In the same exact way, we tend to see our floors visually. So in the same way, can we take a visual approach to solve that problem? Which is, we just use cameras to make and try to detect visually whether it’s a carpet or a rug or hard surfaces. And over time, it just gets better and better along the way. Now there are times when it will get confused—and we do as well—but at that point in time, we can use cleaning head actuation or a few other ways to figure out the difference.

And so there are tricks of the trade over there, but it is really just a vision-based system that we’re working on and solving it that way. That might be—that is also the reason why our robot is slightly taller. One was that the bigger wheels, it can go over thick pile rugs and wires and thresholds and hard surfaces and navigate very clearly. So we just felt like Roombas and Dyson robots were built for a world that was wall-to-wall carpet back in the late nineties and early two thousands. Now it’s much more hard surfaces, thick pile rugs, thresholds, wires.

So bigger wheels allow it to navigate very clearly. But that also means that our cameras are about eight inches high from the floor. So just like a crawling child, it has a top-down vantage point, which allows it to see the floor or obstacles a little bit better than a sort of flat robot. If you have a flat robot, then you might have a sort of ant-like vantage point, which is not necessarily the right way to go and detect various things.

Our Approach to State Estimation

Q9: " How does the team approach the state estimation of the robot? Especially with noise in the dynamic model and sensor data"

Mehul: If you're referring to SLAM specifically, it's a combination of both classical techniques and a neural network-based method. There are two key pieces we've talked about publicly.

One is that we have an image-to-voxel neural net. So what we do is use images to detect depth and turn them into voxels—1 cm3 voxels—and rebuild the entire 3D, Google Street View-like map on the fly. That allows us to create an environment where the robot knows what the distance is. We use long-term SLAM, and as I said, it's a combination of both classical and neural network techniques to actually use it.

And I'm going to brag on behalf of our team here—we do believe that our SLAM is 10x better than anything out there in terms of SOTA or open-source solutions. It's really a spatial AI-based approach that we're taking into account. And really, the honest answer about SLAM is that there are things we've done at a very tactical level that are quite innovative, but in the end, it's just about going and doing it. It’s a grind. There is no silver bullet—it's a bunch of lead bullets we've used to keep fixing every single obstacle along the way to get accuracy to this level.

If you're ever curious, if you have Matic and you're in a room where it mapped itself, just cover the crown—the black part of the robot—with a piece of cloth and move it to another part of the home that it has mapped before. If you remove the cover, it will immediately re-localize itself. And that works in various lighting environments as well. We've worked really hard to make sure Matic understands its surroundings.

Another part we realized is that most robots build what we refer to as relative maps. What I mean by that is they have to start from the dock and then map themselves, so they only know their location relative to the dock. Instead, what we do is build an absolute map, which means that if you've used Matic, you know it comes out of the box and starts mapping. It doesn’t need to start at the dock. The dock is just another piece of information in the map—it’s not a reference point. That increases precision significantly because we don’t have to worry about wheel slipping, dock placement, or any other issues that might affect mapping accuracy.

The analogy I like to use is that typical robots are like, "I can navigate a home if I came through the front door, because the front door is, let’s assume, the dock." But if I came through the side door or back door, I couldn’t navigate. Matic, on the other hand, builds a system that's much more human-like—regardless of whether you enter from the front door, side door, or back door, it can navigate the home because it has an absolute map.

And yes, it’s really awesome that it can navigate even at night. We have IR LED lights on the crown—in the middle as well as on the top—and when we have back cameras, they also have IR LED lights. These infrared lights illuminate the space at night. We use IR because it’s invisible to humans but visible to the robot’s cameras, which capture grayscale images. That allows Matic to clean and navigate with the same precision as during the day. In simple terms, it’s like headlights for the robot—except they’re invisible to humans but fully visible to Matic. And that’s pretty cool.

Advice For Aspiring Entrepreneurs

Q10: "With 7 years of home robotics under your belt, if you could start over what would you do differently? What would you do the same?

Mehul: It's always easy to connect the dots in hindsight. Looking back, we did a lot of research, and of course, there were things we got right and things we got wrong. One thing we got wrong was choosing Umbrella as our SOC platform instead of NVIDIA. That probably cost us about nine months to a year due to delays when we transitioned from that platform. So, in hindsight, I would have picked NVIDIA from the start.

Beyond that, nothing has really changed. Every single thing you see on our website, everything in our product—these are all things we've wanted to have since 2018, even back in 2019 when we did our user research and product planning. Our vision has remained consistent, and so has our execution.

If anything, now is an even better time. We had this inkling back in 2017 that AI would have a huge impact over the next ten years. My co-founder and CEO of Matic, Navneet, worked on Google Coral’s Tensor Processing Unit (TPU), helping develop that chip from Nest’s perspective. So we knew AI chips were coming, compute power would skyrocket, and this combination would enable robotics in a way we hadn’t seen before.

If anything, the future we anticipated in 2017 is now here. It’s an incredibly exciting time to push the boundaries of what robotics can do.

Q11: "What one piece of advice would you give to aspiring entrepreneurs looking to innovate in the field of robotics and AI in 2025 and beyond?

Mehul: For aspiring entrepreneurs looking to innovate in robotics and AI, my biggest piece of advice is this: focus on solving real problems, not just building cool technology.

As engineers, it’s easy to get excited about robots and AI—we love the tech. But people don’t buy robots; they buy solutions to their problems. That was a hard lesson we learned at Flutter. We built gesture detection over webcams, and while it was a cool and exciting technology, we eventually realized that no one wakes up thinking, "Today, I’m going to buy gestures." It wasn’t a product; it was just a neat idea.

When we started Matic, we approached it differently. As a father and homeowner, I want a perpetually clean home, but I don’t have the time to clean the floors myself—and I don’t even want to think about it. That’s the real problem we set out to solve. Robotics and AI were just the means to that end.

So if you’re starting a company in this space, start with a problem. If you get that right, the technology will follow.

Building in America, Protecting Your Privacy

Q12: "Can you elaborate on your pride in building in America as well as focus on privacy?"

Mehul: Having worked at Nest, specifically on the Nest Camera, we realized that cameras inside homes create a different kind of anxiety. Families don’t want to share everything about their lives, and their homes should be a place where they can truly be themselves. Homes are sanctuaries, and introducing cameras or audio/video devices can feel like an invasion of privacy.

In Silicon Valley, while there are many amazing things, we’ve often taken privacy and data for granted, collecting consumer data without proper consent. We wanted to approach things differently. Just like other household appliances work without sending data, there has to be a way to create IoT devices and robots that don’t infringe on privacy.

You shouldn’t have to compromise your family’s privacy just to clean your floors or do your dishes. That’s the approach we adopted—preserving privacy while maintaining a highly effective, low-latency robot. The goal is for the robot to become part of the family, and if it’s a family member, it should respect privacy the same way other family members do (To learn more about Matic's privacy stance)

It’s awesome that everything is stored on-device, and we hope this sets a trend where more developers follow suit.

Shaping Matic's Human-Machine Interaction

Q13: " Mehul, how did your work on gesture and facial recognition at Like.com and Flutter shape your approach to human-machine interaction?"

Mehul: The biggest lesson we learned was that computer vision, AI, and human-computer interaction are means to an end. As humans, most of the time, we don’t want to learn new ways to interact with technology. At Flutter, for example, engineers and researchers used gestures like turning a hand into a pointer, like in Microsoft Kinect, which felt unnatural to us. Point-and-click is an abstraction we invented.

What's more natural is our own body language and spoken language. As computers get smarter, they should understand us without requiring us to learn new ways to interact. This is why we want to simplify things so that we can give commands as easily as we would to a family member. It would be great if you could talk to Matic like you talk to your friends.

We haven’t shipped this feature yet, but it’s something I’m really excited about. Simple voice commands are especially powerful for families with young kids, where your phone might not be nearby, and you’re busy with other tasks. Being able to speak to Matic and tell it what to do is actually really powerful.

We also want to make this technology accessible for the elderly, who may have limited mobility and don’t want to learn new systems. It’s all about making things easy and intuitive.

Is Matic Pursuing Product Placements?

Q14: "Is Matic's marketing team actively pursuing product placement arrangements with any sci-fi shows or movies?"

Mehul: Hopefully, that's the way to go. As the word about Matic and our visual intelligence approach spreads, we’re seeing more opportunities. We’re actively considering various ways to highlight Matic in entertainment. However, once we have features like voice control, gestures, and perhaps mess detection, that’s when things will get even more interesting, and we will definitely pursue opportunities for product placement.

This Was a Marathon, Not a Sprint

Q15: " What was a challenge when expanding your team?"

Mehul: I think we were a little bit lucky in that robotics and AI are exciting fields, so many people have shown interest in working with us. However, the real challenge was ensuring new team members had a product mindset. Ultimately, we want to be a product company, not just a robotics or AI company.

For us, the main thing was making sure everyone understood this was a marathon, not a sprint. It takes a lot of hard work to ship something that just works.

And a lot of it is the analogy we used at the old office: I had a picture on the wall behind my desk that said, "Just keep swimming." So, how do you build a robot that everyone loves? One step at a time. That's been our inspiration for seven years.

Expanding Matic Globally

Q16: " At what point will the Matic robot vacuum be offered in other countries? Is the company waiting for additional investment, or more US users?"

Mehul: Great question! We’d love to do it as soon as possible. For us, the challenge is that we are at the moment barely able to make enough to serve the demand in the U.S. market. With a small team of 60 people, it takes a lot to meet other countries' requirements, from translation and language to different regulations that we have to abide by per country.

So really, the challenge hasn’t been that we don’t want to sell it; it’s just that we have to focus on these particular aspects. And so it’s about realizing those things as we move forward. First, we need to solve it for the U.S. Once the product is complete, we’ll definitely look to expand—hopefully sooner rather than later.

Lessons from Past Startups and New Challenges

Q17: "How has the journey of creating Matic differed from your previous startups? What skills from those experiences did you find most useful?"

Mehul: Great question. I think we knew when we started Matic that we had to build a company and a product that takes advantage of computer vision because that’s where our expertise lies. We were quite aware back then that, since crypto was really hot, if we tried to go into crypto startups, someone would probably kick our behinds—who were experts in that. But our background was in vision, and that’s the one place that has helped a lot.

The second thing is that along the way, we’ve learned what it takes to build a great product—how to build an iconic product and create magical experiences. That’s really the journey we’ve been on: how do you turn technology into a product and experience that solves problems delightfully and simply? And that’s been the question. I don’t want to say we’ve figured it out—we’re far from it. I think Steve Jobs was probably the only one who figured it out.

But we’ve learned along the way how to get there and how to simplify it. That’s really the advantage we have, if there is any, from that perspective. That really comes down to this point of view: I want to have this robot in my home, which is my sanctuary. That means I want it to feel at peace. And noise itself is pollution. If you’re building a cleaning robot, and while it keeps your floor clean, it pollutes your home environment with noise, that doesn’t make sense.

To be entirely honest, we were really worried initially when we started with Matic that maybe with vacuums—because all vacuums are noisy—we just wouldn’t be able to solve it. But as we dug into it, we realized that many vacuum and robotic vacuum companies use noise as a metaphor for efficacy. Just like in the past, we used to believe that the louder the car, the faster it must be. In the same way, many vacuum companies use noise as a metaphor: the louder the vacuum, the more effective it must be, and the more suction it must have. But it turns out that, from a physics and first principles perspective, noise isn’t needed, and higher suction isn’t necessary.

In fact, the Dyson experience I earlier described taught us that higher suction actually damages our rugs. My rug wasn’t old enough to go bald, but using powerful suction for 45 minutes caused part of it to shed. So, we just looked back at the world before vacuums—back in the '60s and '70s when we used sweepers, which were spinning versions without motors that cleaned everything. We realized the best way to pick up dirt is to nudge it, and once you nudge it, you need very little suction to get it into the bin.

Another analogy: imagine a dusty car. If it’s really fine dirt, no matter how fast you drive, it doesn’t get clean. But if you just nudge it with your finger, the dust immediately wipes off. In the same way, we realized that fine dirt requires a brush roll, not just suction. So we spent a lot of time designing a first-of-its-kind hair tangle-free brush roll that handles all pet hairs and long hairs, and it does a really good job of cleaning without requiring huge suction power.

Matic is 55 decibels at optimal efficacy. You can obviously choose a higher setting, but it’s not just optimized to be quieter at 55 decibels. It’s about six decibels lower than a human conversation or your TV volume, which is logarithmic on the scale. So you can easily watch TV or have kids sleeping around it, which is great.

But what we also optimized is making sure the noise is more like a fan noise. Sometimes, vacuum brands like Dyson can make a very high-pitched noise, which is far more annoying. So it’s not only about being quieter—it’s also about the kind of noise it makes. We worked hard at removing every noise that could be annoying to us, kids, or even pets.

The Complexity Challenge With More Sensors

Q18: "Is the problems of complexity of more sensors and actuators greater felt in supply chain or in development?"

Mehul: In both. I don’t think it’s just one or the other. During prototyping, you can probably get one part or sensor, but the integration of multiple sensors is challenging, even during the prototyping and building stages. It only gets more complex as you try to ensure they work reliably, source them, place them correctly, and calibrate them properly.

One thing we did before starting Matic was talk to as many robotics startups and companies as possible. There was a consistent element: many robots don’t work in the field, or their uptime isn’t as expected, simply because, if you have 20 sensors, the probability of one sensor failing every few days is quite high, and that probability adds up.

So, complexity increases, and there are challenges associated with it.

Learning Family Rhythms for Smarter Cleaning

Q19: " Does Matic learn when the house is quite and people are not walking around and that's the time to start working?"

Mehul: Not yet, but that's the direction we want to go. Ideally, Matic would take over the responsibility of cleaning without needing instructions. It should learn that certain areas (like the kitchen or doggy door) get dirty often and clean them multiple times a day. It should also learn when family members are home or not, adapting to those rhythms.

For example, in a home with three kids, like mine, the backyard door is where all kinds of dirt come in, so it should clean that area multiple times a day. Matic could also learn when people are home and avoid cleaning in rooms where we're present. These are optimizations we can make with the vision-based system, and we're pushing forward with that.

Tested for 2+ Years, Upgradable for the Future

Q20: "What is the expected lifetime of the current model?"

Mehul: Great question. We expect it to last at least two years. That's the lifetime we've rigorously tested it for so far, but it should probably last longer. We don't necessarily believe that the lifetime of the robot is limited by the hardware itself.

We would like to keep it going as long as possible, but our assumption and anticipation is that, just like phones and computers, every two years the processing power doubles or quadruples. In the same way, we think that AI chips and AI algorithms will get better as we move forward. So, there's a good chance that by upgrading one NVIDIA chip, we might be able to do a lot more in terms of adding intelligence to the robot. That may become the limiting factor where we may need to upgrade the hardware because the software and AI are progressing so much that processors from two and a half years ago, or four years ago, are no longer the best ones to keep driving it.

And it's honestly so cool to see how Matic has gotten so much better in terms of going under certain cabinets, being able to mop and deep clean. All these new features have been consistently coming out, and the team has been doing a really good job of releasing them. I think we've shipped about 24 app updates and 17 releases since we started shipping production robots. We're in the cadence of doing a minor release every week and a major release every month. We expect to keep that cadence and keep improving super fast.

Validating Assumptions: Cleaning Priorities and the Future Set-And-Forget Experience

Q21: " What is an assumption that you're still validating?"

Mehul: Oh, great question. One of the assumptions we're still validating—and something people are coming around to—is whether making the robot larger, and therefore unable to fit under low furniture, was the right choice. If your furniture is about 12 inches or higher, Matic will go underneath without a problem, but it won’t fit under lower pieces.

From our research and conversations with families, we found that most people weren’t primarily concerned with cleaning under furniture. Instead, they wanted a robot that reliably cleans the obviously accessible areas—like the kitchen and dining areas—where kids and pets are constantly making messes. So, we prioritized daily cleaning of these high-traffic spaces in a more efficient, quiet, and optimized way rather than focusing on getting under furniture.

Under-furniture cleaning is something that can be addressed every two or three weeks. With its extendable cleaning head, Matic can still reach under furniture like an upright vacuum, so it does the job where possible. That was one key design decision we made.

Another major assumption we're validating is the demand for a truly "set and forget" experience—where users don’t have to constantly change bags or add water. Many other companies have approached this by adding a large dock, but we chose not to do that for the first version. We felt that solution wasn’t elegant; in fact, it’s quite the opposite—having a vacuum that needs another vacuum to clean it.

Not to mention, those large docks are extremely loud. If you've ever used a robot vacuum with a dock that empties itself, you know it sounds like a rocket launcher. It always made my daughter cry and freaked out my dog. We didn’t want that. A dock should be elegant and friendly. In our homes, we go out of our way to hide appliances behind cabinets, so why place a giant dock in the living or family room?

So far, our decision seems to be proven right. But if we ever go down the path of creating a "set and forget" experience, we’ll make sure it’s done elegantly—without taking up so much space. It’s also designed to be pet-friendly and kid-friendly. We put a lot of thought into making sure the robot isn’t intimidating—its quiet operation and car-inspired design seem to be working really well so far.