Jake Loosararian, Gecko Robotics

00:00:00.400 | - Gecko Robotics builds wall-climbing robots

00:00:03.320 | and enterprise software to maintain

00:00:05.240 | and protect essential infrastructure.

00:00:07.560 | - He is the CEO and co-founder at Gecko Robotics.

00:00:11.040 | Jake, thanks for joining me today.

00:00:12.840 | - Every hour you save,

00:00:14.080 | you're saving potentially millions of dollars.

00:00:16.920 | - Hardware is hard.

00:00:18.480 | - Be very careful.

00:00:19.680 | - There's a lot of really important problems to solve.

00:00:21.960 | - There's so much sex appeal to building new things,

00:00:25.320 | but you gotta get the business model right.

00:00:26.800 | - The business model has to make a CEO or CFO give up.

00:00:30.040 | - There's a huge fire going on right now

00:00:33.360 | at Philadelphia Energy Solutions.

00:00:35.440 | Oh my gosh, again, look at this guys.

00:00:37.760 | Look at this video right now.

00:00:39.800 | - Today, the Navy remains a formidable fighting force,

00:00:43.440 | but even officers within the service

00:00:45.560 | have questioned its readiness.

00:00:46.960 | - At a missile silo we visited,

00:00:49.160 | time and frigid weather had clearly taken their toll.

00:00:52.760 | - Developing right now, gushing for hours

00:00:55.200 | with no end in sight,

00:00:56.840 | thousands of barrels of crude oil spilling from a tank.

00:01:00.720 | - The report does an estimate of what the need is

00:01:02.840 | to bring the overall grade up to a B,

00:01:05.440 | which is what the society sort of determines to be adequate.

00:01:08.440 | And it's like 4.59 trillion dollars.

00:01:11.800 | - We'll see you next time.

00:01:14.280 | - We'll see you next time.

00:01:15.480 | - We'll see you next time.

00:01:16.320 | - Bye, bye.

00:01:17.320 | - We'll see you next time.

00:01:18.440 | - We'll see you next time.

00:01:19.280 | - We'll see you next time.

00:01:20.160 | - We'll see you next time.

00:01:21.000 | - We'll see you next time.

00:01:21.880 | - We'll see you next time.

00:01:23.000 | - We'll see you next time.

00:01:24.240 | - We'll see you next time.

00:01:25.120 | - We'll see you next time.

00:01:25.880 | - We'll see you next time.

00:01:26.880 | - We'll see you next time.

00:01:28.000 | - We'll see you next time.

00:01:29.000 | - We'll see you next time.

00:01:30.000 | - We'll see you next time.

00:01:31.280 | - We'll see you next time.

00:01:32.440 | - We'll see you next time.

00:01:33.440 | - We'll see you next time.

00:01:34.280 | - We'll see you next time.

00:01:35.440 | - Yeah, all right.

00:02:03.120 | - Hi, I'm Jake, the founder and CEO of Gecko Robotics,

00:02:08.360 | a company that makes robots and software

00:02:11.080 | to help diagnose the health of the built world.

00:02:14.000 | Now, it started in a college dorm, my college dorm,

00:02:18.200 | is now a company that manages over 500,000

00:02:21.000 | of the world's most important

00:02:22.240 | and critical pieces of infrastructure.

00:02:23.920 | Now, the structures that we use to power civilization

00:02:29.440 | have reached their useful life.

00:02:31.720 | It's a huge problem and it's getting way worse.

00:02:34.600 | But it's a problem that you probably don't think about very much,

00:02:36.960 | but you should.

00:02:38.160 | In New York, for example, there are over 17,000 bridges,

00:02:42.840 | most of which are in New York City.

00:02:44.840 | And guess how many of those bridges are not in need

00:02:46.800 | of immediate repairs?

00:02:48.560 | Only six.

00:02:49.800 | See, maintaining things has always been an afterthought.

00:02:55.800 | But that afterthought is now a $4.59 trillion domestic problem.

00:03:01.680 | And by the way, it's getting worse.

00:03:04.800 | It's holding us back.

00:03:06.800 | For example, the military spends 40% of their budget, over $400 billion on maintenance.

00:03:14.120 | Not on building new things, just keeping old things working.

00:03:17.920 | And Fortune 500 companies will lose $1.5 trillion every single year

00:03:25.240 | because of catastrophic failures that were unpredictable.

00:03:28.080 | And our best defense to stop that from happening hasn't changed in over 60 years.

00:03:35.360 | It's this.

00:03:37.280 | This is Joe.

00:03:39.280 | And Joe's on a rope.

00:03:41.360 | Now, Joe's armed with a handheld sensor and what looks like an excruciating wedgie.

00:03:46.800 | Now, Joe's our best chance to ensure that pipelines don't explode,

00:03:54.560 | that bridges don't collapse, that dams don't fail,

00:03:57.600 | and that airplanes don't disassemble mid-flight.

00:03:59.760 | It's an impossible job, unfortunately, for Joe.

00:04:03.040 | You see, we obsess about how software has changed everything for everyone.

00:04:09.120 | It's eating the world, right?

00:04:10.880 | It's important to remember that for the guys behind me,

00:04:14.640 | it's actually never helped them.

00:04:17.040 | You see, the data that we need to prevent catastrophes from happening

00:04:21.200 | in the built world simply doesn't exist.

00:04:23.920 | And without data, what can software do?

00:04:27.280 | Now, I became obsessed with this problem in college.

00:04:31.200 | I was studying electrical engineering,

00:04:32.720 | and my obsession for energy took me to a local power plant in Pennsylvania.

00:04:36.560 | I wanted to see how power was made.

00:04:38.480 | And so, I decided to dive in head first.

00:04:42.160 | No, I actually dove in head first straight through this hole.

00:04:46.080 | And when you got through this hole, you got into a 200-foot-tall steel-tubed box,

00:04:51.520 | the length and the width of a football field.

00:04:53.360 | This is a power plant boiler.

00:04:55.280 | And the boiler's job was to turn water into steam by getting really hot.

00:04:58.960 | See, the problem is, as the plant manager, Jeff, told me, that 40% of the time,

00:05:04.240 | this boiler would be shut down because of pressure tube explosions.

00:05:08.880 | It would cost them $2 million every single day they were down.

00:05:11.440 | This is a small little power plant.

00:05:12.960 | And so, I asked Jeff, "How do you stop this from happening?"

00:05:16.720 | And he says, "Well, we send up humans on ropes looking for invisible defects."

00:05:21.280 | And then he began to tear up.

00:05:24.080 | And he told me a story about how his best friend fell and died the year before

00:05:30.400 | doing one of these inspections.

00:05:32.720 | And he fell and died in the exact spot I was standing.

00:05:34.880 | So, I was floored by the story.

00:05:37.920 | And so, I had to do something about it.

00:05:40.800 | So, I went back to my college dorm and started building the first wall cleaning robot.

00:05:45.120 | And I armed this robot with ultrasonic sensors, just like doctors use for sonograms.

00:05:49.920 | And I deployed that robot into the boiler, saving the plant manager, Jeff,

00:05:54.960 | 30 million dollars just that year.

00:05:56.880 | And I became absolutely obsessed with how we understand the health

00:06:01.840 | of the built structures that we use every single day.

00:06:06.000 | That's why I started a company.

00:06:07.680 | And I boot shopped that company for three years, pouring my life savings into it.

00:06:13.280 | I slept on my best friend's apartment floor.

00:06:15.440 | And I was down to the last hundred dollars.

00:06:18.080 | Two things happened within two weeks.

00:06:20.480 | First, I got an offer to buy the company from a company that makes power plants.

00:06:25.200 | And then second, two partners from a group called Y Combinator said that if I stayed poor

00:06:30.240 | and kept on building the vision, that one day Gecko would change everything that we knew about built

00:06:35.040 | structures.

00:06:35.520 | So, I decided to stay poor and keep building the vision.

00:06:38.800 | And so, we launched the company in 2016.

00:06:41.680 | And we began to deploy the technology into the oil and gas, manufacturing, public infrastructure,

00:06:47.840 | and even defense sectors.

00:06:48.960 | We had to build robots that could climb and traverse and get sensors into all different kinds of surfaces,

00:06:55.120 | geometries, conditions.

00:06:56.640 | And once you have 500,000 assets that you have to climb around, you begin to iterate your robots really,

00:07:02.160 | really well to be able to handle these kinds of environments.

00:07:05.360 | And it became clear that contrary to popular belief from the VCs at the time, the robots were the

00:07:11.280 | mode because they could get sensors to places that could never be gotten to before.

00:07:15.360 | We could convert atoms into bits.

00:07:17.840 | And so, I wanted to double down on that mode.

00:07:20.720 | And so, we started building robots that could fly, swim, crawl, and walk up any surface.

00:07:26.560 | We began to build autonomous platforms to arm those robots to be able to go to places so that

00:07:32.160 | humans didn't have to be in dangerous environments.

00:07:34.080 | We built and became the best in the world at building ultrasonic sensors,

00:07:38.720 | electromagnetic sensors, as well as lasers to be able to see and understand what was going on

00:07:43.840 | inside of steels, composites, and concrete.

00:07:45.760 | We built fixed sensors that could stream live information and data sets to us, both the health,

00:07:52.880 | but also the operational conditions of those assets itself.

00:07:56.880 | And we built an API platform for robots called Fulcrum so that other robotic companies could

00:08:01.920 | actually be used on our platform in streaming data and information live to our customers.

00:08:06.880 | And after 10 years of collecting data on almost every structure imaginable, we launched cantilever.

00:08:12.480 | Our AI and robotics powered operating platform to put those data layers to use for our customers.

00:08:19.360 | You see, when you start building software by first starting out with the data layers and then

00:08:24.800 | building up, you're severely advantaged because you can build software from first principles.

00:08:30.720 | And our ontology now is able to affect folks from the ground level, the guys on the ropes,

00:08:35.600 | all the way through to the executive suite.

00:08:37.200 | It was extremely powerful.

00:08:39.200 | But I was just talking about it.

00:08:41.040 | Let's actually dive into an example.

00:08:42.960 | So to do that, I'm going to take you to Georgia,

00:08:45.120 | to a manufacturing facility that me and you use in the bathroom every single day.

00:08:51.120 | Now this facility has thousands of assets and billions of dollars worth of infrastructure.

00:08:55.520 | And so they wanted us to prove out over 50 assets, what we could actually do.

00:08:59.840 | So I'm going to take you through one of those assets today, a sulfuric acid tank.

00:09:05.520 | So first, what we do is we gather information about the asset by customers sending us their metadata.

00:09:10.400 | And we build out a digital representation of that asset inside of cantilever.

00:09:16.000 | And then we send in our robots.

00:09:18.080 | First, we use a drone.

00:09:20.000 | You can see over here.

00:09:20.720 | Now the drone is armed with cameras that are doing a photogrammetric scan of the asset.

00:09:26.400 | It enriches the asset model itself.

00:09:28.720 | Being able to identify different kinds of defects using point clouds.

00:09:31.840 | So corrosion areas like over here, we're able to categorize and locate.

00:09:36.800 | And then dents and cracks as well, enriching the data asset model.

00:09:40.480 | We want to go further than that.

00:09:42.320 | We incorporate other sorts of components like piping and pumps that you see here,

00:09:46.400 | both in the inlet and outlet.

00:09:47.520 | This is extremely important and valuable because we keep on adding data layers.

00:09:52.480 | And the next one is a dog.

00:09:54.960 | Come here, boy.

00:09:57.680 | Good dog.

00:10:01.680 | Hey.

00:10:03.920 | Everyone say hi, please.

00:10:07.200 | Yeah, there we go.

00:10:08.560 | Nice.

00:10:08.880 | So let me pet him real quick.

00:10:10.880 | He likes to be pet.

00:10:11.600 | Yeah, good boy.

00:10:12.320 | So this dog will walk around to dangerous environments,

00:10:16.800 | gathering information about what's going on on the infrastructure.

00:10:20.160 | Now, what's important also to understand is that because of an API platform for robots,

00:10:25.200 | we've built an extensible way for a company like Anybotics, one of our partners,

00:10:29.440 | to be able to gather information and data sets.

00:10:31.520 | And this robot is extremely exciting because it's built to be explosion proof,

00:10:35.440 | meaning it can go inside of oil and gas facilities and nuclear facilities and beyond.

00:10:39.680 | It's gathering information like you see above, thermal imaging, to understand what's going on with the asset.

00:10:45.520 | All this data is really important when we do optimizations later.

00:10:48.640 | And we want to begin to continue to gather more information about the asset.

00:10:53.120 | So we send in submersible robots.

00:10:55.200 | These submersible robots are looking at the deformation because of the weight of the liquid,

00:11:00.160 | as well as the health of the asset's floor, to prevent things like that oil tank or leaking into rivers.

00:11:06.880 | And once we've gotten this, and the customer is really excited because we can do this while they're online,

00:11:13.280 | we then send in our robots and we rise,

00:11:15.760 | collecting information and data while the customer's tanks are still in operation.

00:11:22.240 | Now these robots that you see right here are armed with ultrasonic sensors,

00:11:25.680 | cameras, and the IMU on board to ensure that we can do this autonomously,

00:11:29.120 | gathering terabytes of data in 12 hours for this tank, a process that used to take about a month

00:11:34.720 | to do while the asset was shut down, costing millions of dollars.

00:11:37.360 | And we can do this in a way as well that ensures that we can localize data points to begin to run

00:11:42.720 | optimizations. And in this case, because these assets are supposed to be reaching their useful life

00:11:48.720 | or reached it, we can extend the useful life of the infrastructure. And so for this example,

00:11:54.080 | we can tell them what to fix in five and ten years to extend the useful life so that this asset

00:11:59.840 | continues to be able to do its function, opposed to having to replace it for eight million dollars,

00:12:05.360 | which was what the plant thought they would have to do. So once we do that predictive model,

00:12:09.360 | we work with maintenance companies to ensure that they actually take action on that,

00:12:13.120 | and we update the model to ensure that a source of truth remains.

00:12:16.320 | Next, the customer actually wants us to begin to do other optimizations, so we use fixed sensors like

00:12:23.840 | this that'll stream information about not just the health, but also the operating condition of the

00:12:31.120 | facility. You see, when you're running a, let's say, a big manufacturing facility, your goal is to

00:12:36.560 | figure out how to make more product without having stuff blow up because of a new operating condition.

00:12:41.840 | Now that's never before been possible because the data you've been able to work with has been from

00:12:46.560 | Joe on a rope. And so you don't know if you change your throughput or make more product, if that'll destroy the

00:12:53.520 | assets, we're able to run optimizations. Now I'm going to show you that here. So this customer was able to,

00:12:59.520 | because we're streaming information and data from the pumps and from the asset itself, we were able to

00:13:04.560 | figure out how to increase the throughput or make more product by about five percent more while only

00:13:10.960 | having to incur over 90 days an accelerated damage of the asset of about two months, equivalent to two

00:13:17.760 | months. And so we've proved that we could do that by actually lowering the fill heights in the tank and

00:13:23.200 | increasing the asset concentration level. You can see the optimization being run right behind me.

00:13:27.200 | Now this was significant because of the ability to not just extend the useful life, but actually

00:13:33.520 | produce more while not having the potential risk of a catastrophic failure, something never before

00:13:38.960 | possible for these companies. Now let's talk about the outcomes for the customer. On average, over the

00:13:44.000 | 50 assets, we extended the useful life by 10 years. This affects their P and L and their margin right away

00:13:49.040 | because of ability to extend your depreciation models. And then we created $105 million of value by being

00:13:55.840 | able to reduce safety risks as well as environmental, as well as being able to reduce the amount of capex

00:14:01.120 | the customers needed to spend. And the estimated from the customer was a four percent impact to their margin.

00:14:05.440 | Now all of this optimization and information coming in doesn't just help the customer, it also helps

00:14:10.960 | cantilever be exceptional in a compounding way at running facilities more efficiently. And so now you

00:14:17.680 | have an ability to have an unfair advantage from companies that are not utilizing technology like this.

00:14:23.040 | So not just are robots cool, but they're actually solving a business problem.

00:14:27.360 | So this has been flying off the shelf, as you can imagine, since we launched cantilever

00:14:32.640 | this year. The 12th largest oil and gas company in the world, for example, determined that they have

00:14:38.000 | 100,000 tanks and that we could provide $122,000 of ROI per tank. Now initially we signed a $30 million

00:14:44.480 | contract, it's exciting, it's going to extend to $100 million, but it shows how if you adopt technology

00:14:50.480 | in a way like this, it's unfair. Now on the defense side, we're working with the Air Force on $130

00:14:59.760 | billion modernization program. Now they have to modernize over 400 nuclear missile silos,

00:15:05.840 | and the best way to determine how to modernize or what the scope was to improve the missile silos

00:15:11.120 | was, I kid you not, Joe on a rope with a hammer, who was listening to the sound that the silos made

00:15:16.640 | when he hit it. So now Gecko is helping to improve what the modernization scope actually should be,

00:15:24.880 | and it points out something interesting. Those that are determining the scope and size

00:15:29.680 | of these modernization projects are the same ones incentivized for that amount of dollars to be

00:15:35.040 | as high as it can be. Now on the Navy side, one of the biggest problems is only a third of our ships

00:15:41.120 | are available to patrol and deter conflict around the world, and the reason why is because of maintenance

00:15:46.480 | cycles. So we worked with the Navy to improve, in this case it was Joe on a skateboard on his belly,

00:15:54.400 | over a flight deck looking at different areas trying to gather information and data sets. We improved

00:16:00.000 | that to be able to reduce labor by 85 percent and improve the turnaround times for flight decks alone

00:16:04.800 | by about a month. So now we're doing tens of millions with the Navy on flight decks and we're extending that

00:16:09.280 | to ballast tanks, hulls, as well as commercial maritime. It's really exciting. And then on the energy

00:16:15.360 | and manufacturing sector that's where most of our seven to eight-year accounts lie with big contracts

00:16:19.120 | like Exxon, BP and beyond. Now one thing that's extremely exciting is that it turns out if pipelines

00:16:27.280 | explode or when oil leaks into rivers it's pretty bad for the environment. So the studies show that by

00:16:35.360 | 2030 in the U.S. you can reduce emissions by about 18 percent if you can stop those kind of things from

00:16:40.560 | happening. So technology is available today to make a drastic impact on net zero. And then it turns out

00:16:49.200 | as well if you're the best in the world of understanding the health of built structures you're actually very

00:16:53.600 | advantaged in building new things and so that's what the admiral in charge of 132 billion dollar nuclear

00:16:59.920 | submarine project called the Columbia class determined. So now we're helping to create the most advanced

00:17:05.120 | submarine in the world from the beginning to the production end. And it gives you a peek into what's coming.

00:17:10.160 | You see, I'm not crazy. Building robotics, material science, AI, software, sensor company, it's really

00:17:22.320 | freaking hard. But I had no choice. You see, the promise of AI from AI companies to make impacts in

00:17:30.480 | these industries have gone empty for years and years and years. And it's no wonder why. They're building their

00:17:37.840 | foundational models off of Joe's data. Data that looks like this. This is a real report from one of our

00:17:48.240 | customers before they used Gecko. It's no wonder that AI hasn't made the impact of the promise that it was

00:17:54.880 | supposed to. So this is why we built Gecko and why I believe because of software being commoditized that

00:18:02.640 | first-order data companies will dominate the next 10 and 20 years in software. And my journey through

00:18:09.040 | the rust has given me both a pragmatism and optimism about the future. A future where understanding how

00:18:14.240 | things work helps you build new things. Understanding how to use AI and robotics in these real practical

00:18:22.160 | ways. A reality where we can understand the health of the built structures all around us just as well as we

00:18:29.040 | understand our own health. And you begin to see robots, of course, in normal society. But these robots

00:18:36.800 | won't be built for doing backflips or folding laundry. They're going to be built to help realize the impact

00:18:44.240 | of AI for the built world with systems like cantilever. Thank you.

00:18:53.280 | David Sachs showed up, everybody. Doggy. Hey. Sachs, that's a robotic dog.

00:19:04.880 | How's your coffee? Your kind of dog. Oh.

00:19:07.520 | Go to Sachs. Go to Sachs. Go give him a kiss.

00:19:14.560 | Oh, there we go. Sachs is very affectionate. You can pet him if you want, David. You can pet him, Sachs.

00:19:19.920 | Here. Where do you pet him exactly? There you go.

00:19:23.920 | That was a lot of love. Yeah. I am experiencing companionship

00:19:29.920 | from this dog. He's excited. Yeah. It's definitely, it's definitely a nice dog. Yeah. Nice dog.

00:19:39.520 | Doesn't bite. Jake, I think one thing that would be great, based on the kinds of customers you have,

00:19:46.000 | can you tell us a little bit about the sales life cycle and the type of deals you do? I mean,

00:19:51.840 | it's so interesting to, is it like an enterprise software type sale? And you know, when you're going

00:19:57.840 | in and doing a physical workplace. I mean, I didn't know where to sit. Were you worried about the dog?

00:20:03.360 | No, this was, it's a very poorly organized conference. Yeah. Let's talk about that.

00:20:08.880 | We only told you where to sit five times in the last four minutes, but.

00:20:12.240 | I don't understand these images. I have, you know. Literally yesterday, we go in there after and like,

00:20:16.640 | Sergei comes, he does his first thing. And I'm like, Sergei's like, oh yeah, do you have any food?

00:20:20.880 | I go out to the food and it's just like rubber conference checking in the VIP speaker area.

00:20:25.600 | And I'm like, Freyberg, can we just get some sushi from Nobu? We should go through all the details.

00:20:30.160 | Yeah. No, tell us about the sales life cycle. So what are the kinds of deals you're not talking

00:20:35.680 | about rubber chicken right now? We've got a panel. But yeah, tell us about the sales life cycle.

00:20:39.520 | The juice has been really good, by the way. Thank you.

00:20:41.120 | Yeah. So the life cycle, it's been, it's been, it's been wild. So, you know,

00:20:45.600 | Gecko actually became profitable in 2017, right after YC's launch in 2016. And so the-

00:20:50.480 | You were a YC company? Yes, 2016. And what was interesting about that was,

00:20:56.480 | you know, we decided to build a company very forward deployed. So instead of building robots

00:21:00.480 | in labs, actually funny story, one of the VCs you had here last year offered a bunch of money at YC

00:21:06.720 | for us not to leave and go back to Pittsburgh and do this forward deployed motion of building robots,

00:21:13.440 | but instead build it in a lab. And I turned that down because I just fundamentally didn't believe in

00:21:17.600 | that way of building. And so, but, so we decided to launch into the, and build robots, like literally

00:21:25.360 | soldered in these environments before, and just figure out how to make the robots work in reality,

00:21:31.280 | in the real world. And so the sales motion was basically, we would go to the plant managers.

00:21:35.840 | Sometimes I'd call and be like, Hey, my pizza guy, like, you know, where's the, where's the plant manager?

00:21:39.440 | Can I talk to him? And I'd get, figure out how to get to the plant managers. And then I'd

00:21:44.080 | convince them to let us work with them in their facilities. And so started out that way by selling

00:21:50.160 | to the folks who need this the most. Yeah. And so, but now I'm talking to obviously CTOs and CFOs

00:21:57.040 | because our products are actually very financial and helps with depreciation models. It helps with

00:22:01.280 | optimizations, but we started by just selling to the folks on the ground and building the robot by

00:22:08.000 | failing a bunch of times there and fixing it live. But now we have a great platform. And so now

00:22:13.840 | when customers buy Gekko, the only way they can buy it is through software. So they buy cantilever

00:22:18.480 | and they bought it, they buy an implementation of the software, which is the robots getting the data.

00:22:22.640 | And then they pay for a license for the software. And we try to make data refreshes,

00:22:27.840 | which is basically robots going out and collecting more information free.

00:22:30.720 | Is there a custom deployment in every one of these? Because they've all got to have different

00:22:34.640 | facilities and how hard is it to kind of customize or you have standard standardization now in each

00:22:40.880 | deployment to kind of do a Chinese menu type selection?

00:22:43.680 | Yeah. We try,

00:22:45.520 | so it's a great question. We started in the beginning by letting the customers pick what kinds

00:22:50.160 | of data layers they want. So data layers basically mean what kind of robots. Now we actually don't

00:22:55.440 | allow them to do that. We follow all the standards, whatever like API, which is like these governing

00:23:00.080 | bodies about how to take care of infrastructure. But then we go way beyond that because I want to create

00:23:04.560 | an incredible user experience that they cannot revert back from.

00:23:07.680 | Jake, there's all kinds of crippling infrastructure problems around the world

00:23:13.360 | that are not necessarily tied to some of the obvious industries like oil and gas.

00:23:17.120 | So I'll give you two examples. One was what happened in Baltimore where,

00:23:20.560 | you know, this, who knows how it happened, but basically the bridge just collapsed in a situation that,

00:23:26.480 | and maybe it was supposed to be, and it did. Another example was a few years ago in Genoa, in Italy,

00:23:32.560 | an entire slab of a bridge just collapsed and it fell on top of an environment and killed a bunch

00:23:39.200 | of innocent civilians. So there's, I think, a public safety requirement here, which is like some

00:23:46.720 | of this stuff was either designed poorly or designed very quickly. How much of that is observable by these

00:23:52.560 | kinds of robots? And how do you convince folks that beyond depreciation and financial motivations,

00:23:59.280 | there's a, you know, a real need to make sure that this public infrastructure is safe and you guys can

00:24:03.440 | secure it?

00:24:04.000 | Um, great question. So the answer to, uh, of how can we actually get information on those types of

00:24:10.560 | instances? Yes. Um, we, we, like you can look at a concrete bridge and say, Hey, there's some decay

00:24:15.760 | here, or there's something that's happening in the girding here. And you can recognize that and learn

00:24:22.720 | and be able to say, wait a minute, you need to send inspectors or shut the bridge down or

00:24:26.320 | stop and figure this out. You first want to do, so what the robots are really good at is getting

00:24:31.120 | a crap ton of data about the assets. And then you can pinpoint exactly where to put fixed sensors

00:24:36.160 | in specific locations that will be indicative, but then also because of our, of our, you know, we have

00:24:41.920 | this like really interesting data set that tells us because of so many different types of situations,

00:24:47.440 | what kinds of potential issues are occurring that we can extrapolate out to these types of situations

00:24:54.480 | that we might be not as familiar with. So we'll put fixed sensors on to give us indications

00:24:58.800 | and give us some ability to help prioritize spending. And so, um, you know, we just actually

00:25:04.080 | signed a contract with governor Shapiro to do this for bridges in Allegheny County and in Pennsylvania,

00:25:09.760 | where Pittsburgh is. And, um, we're helping to modernize bridge maintenance and prioritization of

00:25:16.320 | budget because what you can see here is you don't necessarily need to rebuild stuff. And in some ways

00:25:20.240 | that's not even practical, but you can figure out where to deploy capital. Um, and then, by the way,

00:25:25.600 | did you see this video on X where somebody was going through the Lincoln tunnel and it looked like

00:25:29.760 | it was about to burst? There was like water creaking in and it was really disconcerting. Yeah. Um, but

00:25:36.480 | I think it was more of a design feature to actually like alleviate when times when the water levels were

00:25:41.280 | super high. But my point is there's all of this stuff that we interact with that it would be good to

00:25:45.840 | know that there's a, you know, a service out there looking for it. And is, is there a world where you could

00:25:51.200 | also then theoretically ingest like the actual architectural or CAD of these things and then

00:25:56.160 | also be able to do diffs and variants and be able to tell people, Hey, hold on a second. This is not conforming to

00:26:02.080 | how we thought it should be behaving. Yes. We do. We do pull those in as much as we can, but it's important to remember

00:26:07.920 | that most of the infrastructure that I'm talking about is like 60 years old. Um, now on the new build side,

00:26:13.440 | like for the new Columbia class submarines, for example, um, there's an issue where like,

00:26:18.080 | there's not a digital thread. You have like 5,000 different contractors that are trying to make us

00:26:22.000 | most powerful sub in the world and they're handing paper to each other, um, as they build the submarine.

00:26:27.760 | And so it causes one to be a bunch of delays and issues, which we're seeing with a lot of our ability,

00:26:32.480 | like China, for example, can outbuild us by 232 times, uh, submarines that is, or, or new ships.

00:26:37.440 | And a big part of like art are like, we have to be able to figure out how to be smarter when we

00:26:43.520 | manufacture. And so one of the ways you can do that is digital threads all the way through the

00:26:47.520 | manufacturing process so that we're not like delayed by handing paper to each other that may or may not

00:26:52.880 | be incorrect. And for the customers that we work with, you know, most of them, you know, you're looking

00:26:57.760 | at drawings that are 60 years old. They have never been converted. Um, I, we even try to get asset lists

00:27:03.440 | from customers and they're like, we don't have it. So we have to go out and actually build that for them.

00:27:07.280 | How should we build a submarine just off topic? How should we build it?

00:27:11.200 | Yeah. So we don't have 5,000 contractors at 0.1% or 0.2% the speed of China.

00:27:16.560 | It's a good, it's a good question. I think we should orient to a most efficient, um, way of

00:27:20.400 | building as many components in one place as we can. But you have to remember as well, you know,

00:27:23.920 | congressional members have their own constituents to advocate for. And so they want to bring jobs to

00:27:28.960 | their, to the local communities. And so in a, in a democracy, it's really tough actually.

00:27:33.600 | Mm-hmm. Jake, how do you see, um,

00:27:36.000 | where your bots, your robots, and let's say the more traditional generalized humanoid robots

00:27:42.960 | intersect and when they meet, how do you think about that problem?

00:27:45.280 | Oh, it's a great question. Um, I am so excited to buy as many Optimus robots as possible.

00:27:50.080 | Yeah. You're a customer. You'll be a customer.

00:27:52.560 | A hundred percent. You're not going to be a competitor?

00:27:54.320 | No, no, no. I mean, um, look at any botics right here. So this is a, this is a sweet,

00:27:59.280 | it's a company based in Switzerland that have an incredible robot and their data doesn't know where to

00:28:05.280 | go. And so the, the idea of robots, this is what I firmly believe is that-

00:28:09.200 | Zach's taking them home today.

00:28:10.240 | Is that, um, you know, they get sensors to places that are really hard to get sensors to.

00:28:15.840 | And so that information has to be funneled, um, somewhere to drive some large business outcome.

00:28:21.200 | So I really don't think like, you know, I am not of the belief, I guess, that when Elon talks about,

00:28:26.720 | you know, two times the amount of, um, robots as humans, that you'll see them in society actually,

00:28:32.000 | as much as you may, maybe you'd think, I think they're actually gonna be found mostly in these,

00:28:36.080 | like, really dangerous, behind the scenes, industrial settings, which I, in my opinion,

00:28:41.040 | that's like where they should start for sure, because you'll have to, you know, one, you have

00:28:45.120 | to, those are really complex tasks, but two, they're like very beneficial for humanity.

00:28:48.800 | Um, so like, is there, is there a generalized platform that you've built that allows you to

00:28:53.680 | solve for these different use cases, or do you find that there's

00:28:56.400 | a lot of application specific engineering that's required?

00:29:00.000 | Good question. So we, we, we built an API platform for robots where companies can try

00:29:06.640 | their systems out and we can, because we have a go to market, we can now test if that robot is

00:29:11.200 | producing something valuable from a data side. Right.

00:29:13.840 | And I'm not actually as interested in robots that can weld or robots that can clean right now.

00:29:18.880 | Mostly I'm just interested in just like, what kind of information and data can we build better

00:29:21.760 | operating platforms and systems on? And, um, so that's, that's where I'm starting. And we'll begin to

00:29:27.280 | add more robots that can do different kinds of jobs. But, um, you know, I think it's, you have,

00:29:32.480 | I think this is where first ordered data sets and, and, and software companies have become more and more,

00:29:37.520 | you know, powerful. This is why maybe, you know, um, my opinion on the, the standalone SaaS model is like,

00:29:44.240 | I think it's going away. Um, because the companies that are so advantaged with first order data, you can,

00:29:49.360 | you know, just build your own software.

00:29:51.440 | have a capital markets last question of capital markets kind of embraced the story or things

00:29:56.720 | you can invest. Okay. Yeah. Um, yeah, they, they really are. I think, um,

00:30:03.280 | cause there's a lot of this like hesitation around deep tech and hardware historically,

00:30:07.360 | but you've obviously got an incredible software layer and great recurring business. So you seem to

00:30:11.840 | be pretty differentiated in terms of all, a lot of the long haul build cycles that I think we see out there.

00:30:17.600 | And, uh, yeah, and hardware in our case is very sticky. Yeah. Because, you know, once you convert

00:30:24.080 | from, you know, paper and you're now not using binders, you're using cantilever. Yeah. Um, it's,

00:30:29.200 | it's really hard to go back. Yeah. I just, yeah, I think it's, we've talked before and I think what

00:30:34.480 | you're doing is so inspiring because it feels to me like this technology, you know, that we've seen over

00:30:39.760 | the last 20 years in iPhones in EVs and sensors, um, is now allowing us to move from being reactive

00:30:47.440 | and trying to figure out what happened to then saying, Hey, let's be proactive. What are the

00:30:51.760 | opportunities here to extend life, to sit, you know, extend the life of these assets, um, and avoid tragedies.

00:30:59.440 | And I just, you know, I think the work you're doing is a real grinders work, but it's going to save lives.

00:31:06.000 | And it's going to really save taxpayers money that can be deployed in other places for beautiful things.

00:31:10.640 | And so I just want to commend you on doing something that is so essential. Um, in many

00:31:15.680 | times I meet a founder and they are doing something and I just think, God, I, I don't know if this is

00:31:22.000 | going to work or not, but I know that this founder is going to figure out a way to make it work.

00:31:26.400 | And I think it's just really rare that somebody cares so much about something and then executes as

00:31:32.480 | hard as you have. Um, and I just want to tell you, I personally very much appreciate it because

00:31:36.240 | you know, when a bridge collapses, I know the one in Italy, it's, it was a very big tragedy.

00:31:41.680 | I think it killed 50 odd people and it was privately owned. Yeah. And there's a, you know,

00:31:45.760 | very, very wealthy family that owned it. And ultimately, sadly, no repercussions.

00:31:51.040 | Yeah. Same with dams in Brazil, same with dams in Brazil. Right. So, you know,

00:31:55.360 | as public infrastructure becomes private infrastructure, then the profit motive supersedes the safety motive.

00:32:00.480 | you're going to get all these things unless there's, um, um, some sort of check and balance.

00:32:04.720 | So long way of saying we very much appreciate this hard work that you've done.

00:32:08.000 | Last question. Yeah. You're, you're, you know, you're, you've been in the YC community for a long time.

00:32:12.080 | Just really, you don't have to say yes or no, but have you ever been forced to do any founder mode?

00:32:17.360 | Did you feel pressure? Have you just, just blink twice, Jake, just tell us the truth.

00:32:23.280 | Jake, thank you for joining us.

00:32:25.840 | That's awesome. That was awesome.

00:32:31.360 | That was excellent.

Jake Loosararian, Gecko Robotics | All-In Summit 2024

Chapters