#413 Ahmad Baitalmal: From Tesla to Redefining Warehousing with AI and Robots

Mehmet: [00:00:00] Hello and welcome back to a new episode of the CTO show with Mehmet. Today I'm very pleased joining me live from the bay area. I would say Ahmad Baitalmal co founder and CTO of Mytra. So Ahmed, thank you very much for being with me on the show today The way I love to do it is I keep it to my guests to introduce themselves So tell us a little bit more about you your background your experience and what you are currently up to

Ahmad: All right.

Ahmad: Excellent. Good morning. Thank you very much for having me on this podcast. It's awesome to be here. My background long time ago, I started at Microsoft very long time ago. Did work on enterprise software. Think of, you know, the days of early days of AWS something like that.

Ahmad: Background software in the cloud, essentially. On a large scale then I went to Stanford graduate school of business did a couple of startups after that in home automation sensing IOT, and then I got into Xerox PARC. And Xerox PARC has a long history of innovation and [00:01:00] being ahead of everybody technology.

Ahmad: I was trying to recruit mechanical engineers who are also computer scientists, and I was trying to recruit people from Tesla, and they recruited me back. I said, no, we've got this little project called the Model 3 Ramp, and we think you'd be great at it. And so I joined Tesla right as we were doing the ramp for a model 3 which was a great experience.

Ahmad: Learned a lot about manufacturing and was able to help out with the ramp in 2018 for the model 3. And then the bring up of the manufacturing, uh, factory in Shanghai and then Berlin and Texas. And then I was recruited into Rivian to pretty much do similar things.

Ahmad: So I was director of factory systems at Rivian as well. And I thought I was done after that. I was in Hawaii just enjoying my time. And then one of the people that I really had a great time working with at Tesla Chris Walty. We started talking about the pain of manufacturing you know, [00:02:00] the things that we were dealing with over and over.

Ahmad: And he kept saying, you know, we can do better. And I'm like, yes, we can absolutely do better. And it's one of the best things I learned at Tesla. It's that you really can apply a lot of the efficiencies that we learned from cloud computing and fault tolerance, all to the physical world, and there's a great opportunity to do that, and that's what.

Ahmad: Let us to start Mytra. We thought that we can rely on our engineering background and solve one of the biggest problems right now in logistics. And so that's a quick rundown of, You know, my background and how I got here.

Mehmet: Fantastic. And it's been really fascinating also I'm at like, and very rich.

Mehmet: So if I want to start like, you know, because you know, like, you have also like the background that I was preparing for the show as a pilot, right? So this discipline of aviation and, you know, having the safety protocols, decision making, you know, how it influenced also [00:03:00] your way in technology leadership and Product development currently at Mytra.

Ahmad: It's really fascinating. I'm really impressed with the aviation history, especially here in the U S it's been developed over many decades. And there's a lot of learning that you go through just simple disciplines like checklists. Things that's, you know, you have to go through every time and standard procedures if you lose an engine in flight, here's what you do.

Ahmad: There's a quick way to just go with the, you know, what's called a flow, which is things that you have in memory, like muscle memory, where you just apply it and then you back it up with a checklist, did I miss anything? And so having these things prepared ahead of time. You know, applying it to your production in your production.

Ahmad: It's really like, you know, taking flight. You're saying I need to produce or I need to, you know, accomplish a few tasks here for the company is very critical for the company to to [00:04:00] accomplish these things applied to something, a company like Tesla. They say we are going to produce the model Y, you know, announcement sometime in March, 2019, by the next year, it's like less than 10 months production starts, this is different than any other company.

Ahmad: And so to do that, you're saying, okay, what do I need to do? Let's line things up. Here's the checklist of things that we have to prepare. And of course things will go wrong. It's not like maybe it'll go wrong. It'll absolutely go wrong. And so you prepare ahead of time, you do your training. You Go through what could possibly go wrong and how do you mitigate?

Ahmad: And so at the time when something goes wrong, you're not figuring out on the fly. You've already got your decision tree done for you. And it's almost like just executing these things help you land, you know, the plane or your production safely, I thought I found a lot of analogies between the two disciplines.

Ahmad: It also prepares [00:05:00] you for dealing with ambiguity as a leader especially in the technical uh, industry. There's a lot of unknowns and there's a lot of things that you might, it might work. It might not work. You just don't know. And so you have to practice dealing with ambiguous situations.

Ahmad: You are looked for guidance by people in the company. They don't know if investing in AI is really important or investing in mechanical, new mechanical innovations is important. And you are providing guidance, and so you have to, at the very least prepare. Look ahead, see what is the strategic best approach and see how you can land this safely.

Ahmad: Yeah I hope that answers your

Mehmet: question. Yeah, absolutely. And you know, like also I think Ahmed, one of the things which I'm sure like also you can You know share better than me I'm also a fan of aviation [00:06:00] also as well, you know, and you know all these I would say behind the scenes thing that happened because you have to Deal with multiple people and you need like also to manage the resources properly Relating this of being you know in a leadership position As a cto or any other leadership position in technology.

Mehmet: So I think this also like Kind of there's some there are like some similarities because technology today In any sector I would say it became like paramount and you know, like every single You know small problem can have huge impacts. But what i'm interested in is like how also this You know, like shape the way of your leadership style.

Mehmet: I mean, building like high performing teams, you know, training them on having the resilience and having, you know, this ability. So if you can share with us like a little bit about this aspect from leadership perspective.

Ahmad: Well, it's like being a [00:07:00] pilot in command. So you're in command of this this venture that you're leading.

Ahmad: Of course you're human. You cannot, Say I'm doing everything. And at least at Mytra, it's one of the key values that we have. We say we're not experts, and so, we hire people who are experts and we rely on them on making decisions that are highly impactful to the company. I don't know everything about.

Ahmad: Electrical engineering, for example, or how to wire specific motors, you know, there are people who are better than me at doing this. And so we go and we try to recruit you know, these people but the important thing is that we're aligned 1st. So, if we say we're going South. Here is the direction itself.

Ahmad: That's our direction. As long as we're all pointed at the same way, I will rely on the expertise of the, you know, the people that we hire and they'll make decisions that I [00:08:00] don't have the ability to vet myself. They could say, are we going to use a different kind of battery? 48 volt system instead of 24.

Ahmad: I cannot argue. I cannot you know, I will depend on them making the right choice for the company. And so what's really important at that point is that we're aligned that we see the same future and that becomes my job. me, the person who is, you know, setting the North Star. Let's define that.

Ahmad: Let's say, you know, convince people. Here's why we're doing it. Here's why we're doing it this way. Could you please use your expertise to help us get there? And we do that across the company. So everybody in the company as they are being recruited and when they first start, We have these conversations or say, here's why this company started.

Ahmad: Here's how we need to do it. The decisions that they make at this stage are extremely critical. You can [00:09:00] steer the whole company in a completely different direction. So that becomes very important. Yeah, it's a very big part of the Mytra culture the alignment

Mehmet: part. Fantastic. Ahmed, let's dive a little bit into, you know, what Mytra do today.

Mehmet: I mean, and, you know, the, some of the challenges I would say, because when we talk about and again, like I'm a little bit biased because you know, anything that involves manufacturing machines and I'm, you know, a computer engineer by trade, but again, like I was always, you know, attracted to the mechanical part and the manufacturing part and so on.

Mehmet: So what is the main goal, you know, for starting Mytra and you know, what are like some of the problems That traditionally, you know, your ultimate users, I would say they would have and how you solve it to them.

Ahmad: Okay, this is a very good question. This is something that is very critical for us.

Ahmad: If you are [00:10:00] trying to, well, if you look at manufacturing as general manufacturing warehouse management, they, they really look alike. And so. If you're trying to do any kind of industrial operation, you're bringing material in, you're transforming that material, adding value somehow, and then you're selling it that 95 percent of what really happens is the movement of this material to be processed and your options right now globally.

Ahmad: Your options are limited. Almost everything that you see around you, if you look around the room, almost everything has been on a pallet. And a pallet here looks like a wooden or plastic piece of square structure that gets carried by a forklift. This is 50 inch by 50 inch. We don't notice these things.

Ahmad: But now you'll start noticing them. You'll see them at your retail stores and hospitals and warehouses and airports. They are everywhere. They move material. I call it [00:11:00] the red blood cell of the global economy. Everything moves on a pallet. Now, your options. To move this pallet are generally fall into two buckets manual, meaning a human gets some kind of tool.

Ahmad: What's called the pallet jack or a or by hand, they put it on some wheel system and they push it around. That's how material moves around the globe. The other option is to use a forklift and we've all seen forklifts. They are these you know, driven vehicles in a warehouse and they have two spears that go into a pallet and they move it.

Ahmad: Now your refrigerators, your paint, the stuff that you buy from Ikea, everything has been on a pallet. And so those are, these are the two options there. And we noticed that in manufacturing, we saw that material come, the car parts, Arrive on pallets. They come out of the trailer and somebody manually moves it into the production line.

Ahmad: And that is where complication [00:12:00] starts to happen. There are multiple systems that get involved. You first have to buffer these things and store them in storage. That's where warehousing comes on, comes in. And then that material has to selectively be, you know, Taken out of storage and put into the production line and multiple systems, touch it conveyors and lifts and tracking systems.

Ahmad: Each one of those systems have been provided by a vendor. A car manufacturer really doesn't produce the storage system for their warehouse. They don't produce or they're, they'll not manage their conveyors. They don't they hire people, other companies to do it. And that applies to any kind of. Industrial operation.

Ahmad: You want material to move from one end to the other. You have to hire a vendor to create a, you know, some kind of system to move it. And we saw that as the most painful thing that we dealt with. And it was so, prevalent and people just stopped thinking about it. They said, this is it. For me to know where [00:13:00] my car parts are, I have to talk to the vendor, talk to their, you know, find out what database they use, how they integrate with the next vendor that takes material from one end to the other, and Each one of those vendors, their job or their incentive is not to produce the thing that you're trying to produce.

Ahmad: Their incentive is to get the next contract and that's it. And so we were trying to build cars and if something was missing, I'd have to go and talk to each one of these vendors. Like, Hey, the bumpers didn't show up. What's happening. RFID readers, lifts and conveyors. It's just too complicated. So we sat down and said, okay, what would be the forget about what's available now?

Ahmad: Let's just start from scratch. What needs to happen? We started with the standard weight of a pallet Now there are systems that move totes and bins and you know small boxes for groceries and stuff But we said no really we want to focus on [00:14:00] you know the hard part is to move a pallet that's 3, 000 pounds the general max weight of a pallet, if you were to put liquid on it, you have to support about 2, 250 pounds, but so we targeted 3, 000.

Ahmad: That means I can put refrigerators or part car parts on it, or, you know, tires, things that are heavy. And we said, fine, let that is the unit of work is to move 3, 000 pounds freely in a warehouse. And we started working backwards from there. What kind of thing I need, do I need to build today? If I want it to move 3000 pounds on a pallet, as I said, manual or a forklift to carry 3000 pounds.

Ahmad: The forklift itself needs to weigh 9, 000 pounds and that's what they are. And they cost about 50, 000 and there's an operator on top of that. So that's the basic unit of work to move 3, 000 pounds around. You need a 9, 000 pound pallet and an operator to [00:15:00] operate it. And we thought. Let's build something to do the same work.

Ahmad: And so that's what we ended up with. We spent close to nine months figuring out a way to build a robotic system to move these 3, 000 pounds. Our bots weigh about 500 pounds, not 9, 000. So that changes the math completely. They are autonomous robots. They don't need an operator to move them. They don't need aisles to move within the warehouse.

Ahmad: And so that solves a lot of, it just changes everything and it makes it. So your warehouse is software defined, meaning that I have a big grid. That's why it's called Mytra from matrix in Greek. It's a grid. It's almost like Minecraft and the bots move within that grid freely. Freely. They move 3000 pound pallets and full 3d left, right.

Ahmad: Backwards and forwards and up and down and you could just say, pick it [00:16:00] up from this cell to that and drop it up in the other cell. And so you allow the warehouse to be completely controlled by software. This sounds like a dream. This sounds like something futuristic. And so that was, we said, hey, this is a hard problem.

Ahmad: Nobody's tackling it head on and we decided let's do it. So, it was exciting. That's what got me off the beach at Hawaii. This problem.

Mehmet: Yeah. I advise everyone to go and have a look at your website. Of course, I got to ask you about it at the end because it's really futuristic, you know, seeing these boats able to move, you know, the pallets from one place to another place in an autonomous way.

Mehmet: Okay. And it looks like a merge between, you know, the art of robotics and mechatronics, if maybe I'm using the right term with, you know, all the beauty of, you know, emerging technologies that's coming. Like I'm sure like you have [00:17:00] implemented some machine learning also as well into the process.

Mehmet: So, so I like the technology. We don't have to dive too much technical, but on a high level, you know, like what, like are the components of having, you know, this, you know, I would say it's state of art. Seriously, Ahmed,

Ahmad: it is a mixture of all disciplines that are required. So we have to figure out you know, mechanical innovation electrical, and software and you know, logistics as well.

Ahmad: And so you have and structural structural engineering. So if you go to the site and you see how it looks like, it kind of gives you an idea and you think, okay, what's involved in doing this? Well, first of all, we have to figure out a way. To move 3, 000 up and down in this structure behind me.

Ahmad: Um, the tough part was to reach out to experts in mechanical engineering and say, hey, I'm think, I really need to do this. How [00:18:00] can I do it? You get a lot of pushback like that. You don't need to get something to move or shuttle the pallet, you know, on the floor and then have it go into a lift and then the lift goes up and down.

Ahmad: And so that's another component that you have to add to your system. And that's, that comes from a vendor or if you're to bring it in, that means I have to hire a team to work on the bot and a team to work on the lifts and that's just the wrong approach. So we wanted to make sure that it's the one device, the bot and the structure.

Ahmad: So we had to figure out a way to do it. And mechanically, we spent an iterative, we had an iterative process to figure out how do you build a thing that can go up and down and carry 3, 000. And so, the thing we ended up with is a helical drive and you'll see it on the website at the four corners of the bot.

Ahmad: There are these arms that extend out. And each one of the arms has a gear that is helical. It is, it looks like a helix and it has [00:19:00] wheels on it. So you get both benefits. You have low friction as the arms turn. It's almost like, four screws on the corners and it has high torque. And this was a process where we start off with.

Ahmad: Very high friction, very high heat. You know, as you're, you know, trying to screw this thing up and down, but then we kept figuring out ways to solve the next problem. This was just a mechanical, you know, innovation that we have to invest in. And it was a little bit scary. You go like, I'm not a mechanical engineer.

Ahmad: I have to rely on people that tell me, yes, this is doable. No, it's not doable. So as long as you don't break the laws of physics let's just do it. And so we did, and now we have a bot that can move up and down in full 3d. That meant also. We rely on our EV experience. So we have you know, some of the best people that worked at Tesla for, um, battery technology and motors to help us build this [00:20:00] essentially an EV inside the structure.

Ahmad: That's the electrical part, a lot of the software that's controlling this. Now you think, Oh, it's just. You know, a bot with a bunch of motors, you just turn on four motors and you go up and down, well, you're managing 3000 pounds and they're shifting. There could be liquid, things get moved and there are different frictions for every corner.

Ahmad: There is a kinematic model. So one of our leading roboticists here Dr. Elena Galbally, she's a Stanford PhD in robotics, and that's her work at was to figure out a kinematic model to balance. You know, that level of weight in a robot that's climbing up and down. This is purely software mechatronics.

Ahmad: It's a mixture of like firmware. There's a lot of work that had to go in there. We're using NVIDIA GPUs and TI GPUs on the bot to manage all that. The bots, there are multiple bots that work on in the structure at the same time. So [00:21:00] there's eight cameras. Now it's going to be 16 cameras.

Ahmad: There's a lot of computer vision that's in there for them to see the material, see other bots. I, it's a mixture of all disciplines and we're trying to get the best people in the industry to kind of come and solve their part in that puzzle. I'm still surprised and amazed when I see the bots moving.

Ahmad: It just looks so futuristic to me.

Mehmet: It's absolutely amazing. And you know, like there's a term you use, which is like software defined warehouse, right? So, so this is by itself, it's like, it's a revolution because if someone tells me, okay, there is a warehouse, like I can't imagine, okay, how I can bring technologies there.

Mehmet: Yeah, of course I can put sensors. I can think about, you know, you know, Few things here and there that can do, but you know, you, you have really had a revolution in the way how the inventory in the warehouse works. And yeah, all of us have watched, you know, documentaries seeing these you know, the, what you call them the ones that they [00:22:00] move you know, these machines that they move the pallets.

Mehmet: Yeah, the forklifts. Exactly. Yeah. So, so, yeah. So people, they see, of course, the, you know, the guy who's, you know, like just driving these ones, but now I'm seeing bots just out of curiosity. Does these bots can in somehow communicate with each other or coordinate with each other?

Ahmad: Yes, absolutely. So that's a big part of what we're doing.

Ahmad: And it's a phased approach meaning that one critical thing I learned from my time at Tesla, and everybody could see that if you drive a Tesla, like I bought one in 2018, it's the same car that I have today, but the features have improved significantly since the time I bought it because the hardware was already there.

Ahmad: And it was designed so you know, you can update and add features just with software. And so cameras have a big role to play. Where we are taking the same approach. We are putting hardware in there. For the bots to start learning about [00:23:00] the features that we're or the tasks that it needs to do, and then extend that feature later on just with software.

Ahmad: A big part of it is coordinating among the bots. And so right now, the bots are able to talk to each other. There's a central point where they talk, but the next step is to enable them to talk directly to each other. One, one, one bot to the other. And that is a big critical differentiator for Mytra.

Ahmad: The bots have got to be autonomous. By that, I mean, we have to drive the reliability and scalability of this system. Big words, but let me explain. The biggest problem with automation in general in warehousing or in any industrial setup is the reliability part and scalability, it sounds great on paper when the contract is being signed at the beginning, Hey, we're going to automate and it's going to be great.

Ahmad: And the reality is things break all the time. And so, for me, [00:24:00] experience, a very positive experience that I had working at Tesla was making every station independent, meaning that it has enough compute in there. For it to operate without needing to connect to, you know, a central brain in the cloud. And so that enabled Tesla to scale really fast.

Ahmad: People think, oh, Tesla will do X number of cars. They surprise everybody at the end of the quarter, they're producing a lot more. That's because of the fault tolerant approach that Tesla has in their manufacturing software. It was great experience. And so we're applying the same thing here. The bots have got to be autonomous, meaning that they need to handle, you know, simple exceptions on their own.

Ahmad: And to do that, they need to communicate with each other. So every bot, The analogy is a manufacturing station or some processing station in a warehouse has to be able to independent, independently operate and then pass on that information to the next station or broadcast it in that group. [00:25:00] That means that there is no single point of failure for the whole system to go down.

Ahmad: But you have to build that in the device, on the device itself. So a big part of our bots the big part of our bots feature set is they are absolutely able to talk to each other. They're able to work to unblock things. So think of this big grid of material and there's a palette that's in the middle, but it's blocked by other material.

Ahmad: When a bot arrives there to pick it up, it'll notice, Hey, I can't get this out. I need help. Either it does it by itself. Or a faster way is to communicate with the other bots around, like, Hey, if you're free, move this out of the way and let me take this up. Those things fundamentally are not hard things. I mean, logically it's very easy to build a you know, logic to do that.

Ahmad: If you are human and two people are coming in to carry [00:26:00] something out and one of them is blocked, you talk to the person next to you and say, Hey, help me out, get this out of the way for me so I can take it up. It's easy. The thing is automation as it's done today does not take that into account. So companies that adopt automation cannot scale.

Ahmad: And we're saying, no, the bots have got to have enough compute and enough ability to communicate with each other. Okay. So you as an operator say, I want this to happen. I want this material to come out or take this material and store it. And the boss should, they should be able to handle it. They have enough intelligence and enough ability to communicate and coordinate amongst each other that they solve the problem.

Ahmad: And when you have a system that is self solving like that, scalability becomes Possible, you know, whether you're running a small warehouse or a million square foot warehouse. It does not matter because it just works. And so that's what [00:27:00] we're aiming for. That's why we're building my truck this way.

Ahmad: I do not want a a warehouse system that has a single point of failure that cannot scale. We want to make it so. You deploy a micro system and the incentives for us are the bots handle it without needing intervention as much as possible. So that, that's why this, Communication between bots is critical and there are many ways to do it.

Ahmad: There are many technologies out there to help. Plus if you're relying on communication to happen with a, your standard wifi network or LTE network in a warehouse, that means your whole operation is dependent on this tiny thread called wireless communication. But that's not really, it's really, that, that is too risky.

Ahmad: And for us, it was like, no, let's get the bots, To communicate with each other. Two bots standing next to each other. They should be able. We have the technology to have two [00:28:00] machines talk to each other and say, Hey, let's coordinate this movement. It's a simple movement. There are things that are too complicated.

Ahmad: But for things like that, we should be able to do that.

Mehmet: Yes, fantastic. You know, like the way I also described about, you know, how it should be like resilient and you know, there is no single point of failure. So it's kind of you have, you are building maybe I'm not sure if the term would be appropriate.

Mehmet: Be applicable, but coming, you know, myself from infrastructure, background and technology. So it seems like you're building a private cloud for the warehouse. Like, you know, like, so you want all the components to be like always highly available. You want like no single point of failure, like uptime 99. 99%, which is fascinating.

Mehmet: Now mentioning the AI, Ahmed, and You know, of course, like everyone talks about the AI and, you know, having yourself, you know, the background in usually optimizing, you know, how stuff works. Do you see, you know, something which untouched [00:29:00] still when it comes to AI? I'm not sure if generative AI will apply here, honestly, but like how, what are like some of the, you know, Areas where do you think still they I didn't, you know, being utilized in that can reshape, you know, the supply chain and the physical operations.

Mehmet: And, you know, maybe we don't have the attention to that yet.

Ahmad: It's actually going everywhere to some degree. The areas that I think are not being To my satisfaction, at least really exploited is the idea of exception handling think of your basically why do we rely on humans? So if 95 percent of the warehouses are still using humans, you go, why, you know, since the eighties, when automation started to come into manufacturing, people thought, oh my God, the robots are going to take over.

Ahmad: It has not. Robots did not take over. And the reason is that humans are the best programmable machines [00:30:00] that we have out there. You do not need to program a human that much. You could tell them, Hey, perform this task. And if something, an exception happens, a box falls on the floor. You don't have to go and program a human to pick up the box and move on.

Ahmad: No, they've, they know that already. And so that's a big difference between humans and machines. And now with AI. I see a lot of attention given to generative AI, transformers the transformer purchase is great. Yes, it's a predictive model. It says it was all based on. What's the next word based on the context that I have so there are applications there What I think the next opportunity for AI is the exception handling part meaning that you deploy machines automation and Something goes wrong because you're never going to be on happy path all the time And the amount of interventions that are needed by human prevents companies from relying more on [00:31:00] automation, more on it to the point where they're not going to scale.

Ahmad: They say, this is, this requires too much support. And that's where the cost comes and it becomes super expensive to support a system like that, that they just don't deploy any more automation. That's why we don't have the automation revolution that, you know, we were expecting. And so where AI can help next is to do the exception handling.

Ahmad: The example I gave where two robots are, you know, trying to perform a task and they need to collaborate amongst each other. It used to be something that had to happen in some number one, you had to know that it's going to happen, plan for it. Program manually, the steps on how to resolve this problem. And then deploy it.

Ahmad: It's complicated. If I was running a business, say a bakery, I'm not going to hire a team of AI people or software engineers to figure out the side, you just rely on your vendor. And the capabilities of the vendor is [00:32:00] transfers to me by, you know, acquiring their services. What AI, the area, the next area for AI to help with is this ability for robots or automation.

Ahmad: Or any industrial machinery on the floor to solve these things by itself, to start doing the exception handling that you would expect a human to do without you needing to predict what the problem could be and figure out ways around it and pre program it manually. This is very promising and I don't think it's given enough attention.

Ahmad: At Mytra, that's exactly what we're trying to do. A very good description is you could tell the robots, Hey, here's all the material in the warehouse. Here's what the material is. Here's where every other robot is. Here's the lighting condition. Here's the temperature. This is like prompt. In general, you're saying, here's the context.

Ahmad: And then you ask the question [00:33:00] says, What's the next move? That's exactly what you're doing. That's how generative AI can help. And the next move for the robot would be, Oh, I have this task. I'm going to do this based on all the context that I have. Maybe there are other companies that are secretly working on this, but I see a lot of attention given to conversational AI generating videos in the industrial setting.

Ahmad: There's a huge amount of, a large amount of value that can be derived from general AI to make these decisions at the bot level. And this does not require huge compute in the cloud. This can run this inference compute. That's required to make these decisions is on the device. This is where a lot of value can be unlocked with AI in the industrial automation, this on, on, on device inference compute, that is the next revolution in my opinion.

Mehmet: And I think, you know, the technology is allowing especially to run, for example, the LLMs [00:34:00] locally on, on the local device itself without relying, you know, on going to something like open AI, chat, GPT, or like, Gemini and so on. So, so I think, yeah, the technology is also. Pushing to that direction of having this every device can run its own LLM in a way and, you know, be able to be autonomous.

Mehmet: So, that's how

Ahmad: the whole smart devices thing, you know, a long time ago, smart devices were really connected devices. They weren't really smart. They were just connected devices. Yeah, I have truly smart devices.

Mehmet: Absolutely funny enough, like my father used to tell me, I don't know why they call them smart, like they don't do anything.

Mehmet: They just reply to what, I mean, they just act as I ask the screen to scroll up, scroll down. So there's no, no smartness in this. Yeah. So one thing, you know, we are talking something very futuristic actually Ahmed with you and, you know, for you, you have a very long experience and you've seen like multiple initiatives and [00:35:00] you let yourself multiple initiatives for these kinds of innovations.

Mehmet: So how do you do the balance between, you know, something that is really reasonable to do and achievable versus, you know, Like something which is, I don't know, like it might need like more time and more more efforts. So how you manage to stay practical, I would say like balancing moonshots, let's say, versus like something which is practical, immediate, that can be done very fast.

Ahmad: I'm

Ahmad: trying to think about the best way to, to answer this. The, if it was something easy, you know, most people. We'll just do it. You know, you're not adding a lot of value there. It's a big part why we decided to do the bots this way. It's harder. And but you have to weigh it with the expected [00:36:00] value created out of it.

Ahmad: Number one, you have to start with, are you breaking the laws of physics or math or statistics? And you go, okay, let's get that out of the way. Let's just make sure, is this even possible? Is there something stopping us from doing that? If it is possible, then you go, the next step is this really solving a problem that, that has some value, some economic value.

Ahmad: And so we looked at moving the bot in full 3d. Yes, there are solutions where. A pallet can be lifted up using an elevator and there are conveyors out there, but you see the complicated integration points and you see how that's hindering the adoption of these things. And you go, is it worth it?

Ahmad: And that's one of the first things we did. Is. simulate the whole system, simulate bots moving with the motion profile. And the good news is you could do it with software. Now you don't have to go and actually build it, but you can simulate a whole system where pallets move in full 3d without the [00:37:00] need for aisles or forklift.

Ahmad: Pathways. And it's, you use 100 percent of your storage square footage and say, if I build something like this and it moves in every direction and I have, you know, 10, 20, 30 robots moving. Let's simulate 10 trailers going in and 10 trailers leaving. What's the efficiency, what's the efficiency, how much energy did I use?

Ahmad: All these things. Then you go, wow, this is, you know, 10, 20 times better. So you see the economic value, you see that the technical aspects of it are doable. I mean, you're not breaking the laws of physics. It is something achievable. And it's, you know, sometimes really great to have. You know, incumbent experts tell you, Oh, don't do it.

Ahmad: It's if it was doable, so and so big company would do a wide one. Just like, great. That's a signal. That's great. I'm going to go for it because there are no incentives in the industry to kind of improve. So you see these [00:38:00] little signals and you go, okay if that's the case, let's tackle it. Now, the heart of the problem.

Ahmad: The more reward is expected. It's a very classic thing. You know, high risk, high reward. It's easy to to understand. I have another analogy that I used with a lot of people recruiting who said, Hey, if If this operation was like, if you wanted something safe, you can just invest in a gas station.

Ahmad: Gas station economics are pretty known. Here's how much gas costs or to sell. Here's a location in a real estate location. You can draw out for the next 10, 20 years. Here's the expected output of that. It's very low risk, well understood economic project. that you can do. But if you wanted to go somewhere a little higher, you want to create more economic value, then you need to approach the hard things, the [00:39:00] things that people shy away from.

Ahmad: And so that's a very good, rich, fertile ground to create value. Is there something that is extremely hard to do? Unknown ambiguous, but the promise of it actually working. Is very valuable. You start looking for the right people that have that ambition, that have the expertise to say, Hey, you know, this is doable.

Ahmad: You know how to do, you know, mechanical or innovate in mechanical or electrical or, you know, in the software, can we all get together and see this attractive, you know, goal north star in the future we think is doable. And so you kind of study it and then, you know, draw a path there and start executing.

Ahmad: Obviously there are risks, but that's the whole point. You take the risk and the reward is hopefully amazing.

Mehmet: Two, two answers. I like them a lot. And you mentioned one of them, which is like, yeah, [00:40:00] don't even try to do it. The other one, Oh, we have done it this way all the time. Yes. Right. And, you know, like, for me, even like in internet, like maybe starting a project, but it always triggers me, Oh, let me go find why these guys are saying this and then what I find out, like, even if maybe one of the guys is right, like it's not doable.

Mehmet: But what I discovered, like, by trying, I learned a lot of other things on the way. And they say, okay, why they didn't try it? Like, and, you know, this brings to me also like the first principle, right? So, so we always, if people tell you, oh, we have always done it this way. Okay. Let's see if there's another way.

Mehmet: Maybe there's not. But it's worth a try because at least we can put it on the book. Okay. Don't try this because we did it and it didn't work for us better than just saying, okay, you know what? Like, no, don't do it. So absolutely. I like this now mentioning this and I will, I don't know if we're coming to an end, but you know, I'm enjoying the conversation with you, Ahmed, the role of the CTO, like with all.

Mehmet: These things that we just mentioned, like, and all the technology that started to emerge how are you seeing the role of the CTO [00:41:00] also emerging with all the, you know, not only the generative AI is one of the things that came, you know, recently to the spot it's been for a while, but now we are having like kind of this flood of new emerging technologies.

Mehmet: And as a CTO, you know, how should I, You know, act. What should I do to uplift my game? What you can advise, you know, also like fellow CTOs or people who are on this career path, I would say.

Ahmad: Yeah. So, I thought about that a lot and I noticed there's a shift that happened in the past. Maybe let's say 15 years.

Ahmad: Early on there was a focus on software, which is great. I mean, I agree. My background is software engineering. I love it. I think it's extremely valuable because of that. There was a push, at least here in the U S to push manufacturing, the physical. Manufacturing stuff outside [00:42:00] offshore. And it created this mood in the CTO, you know, cohort where they say, Hey, let's do the fast, fail fail, fast pivot, all that stuff.

Ahmad: Great. Yes. Yes, that I agree. Those are proper ways. But if that becomes the prevalent way of. Operating and it takes over, then you've lost something you've lost the the drive to take larger risks because if it's just software, you're wasting, you're saying, Hey, this has low cost of, you know, entry into the market.

Ahmad: Let's try to do another SAS app or another social network. And the market gets saturated by these things. And all the people who were CTOs in that area think in those terms. You know, try something that's frivolous. Don't go for the hard stuff. Go for something that has huge potential margins right away.

Ahmad: Like social network, if it picks [00:43:00] up great, the returns are amazing. And so you're just doing these quick things back and forth. And what happened was a move away from taking bigger risks, especially on hardware. Hardware has always been looked at as high that's required that I need to, you know, it's very hard to develop something hardware, very hard to iterate on hardware.

Ahmad: It's much easier to do things in software, correct? Yes. But now the mode has been, let's just do the easy part. No, let's focus on the hard part. You as a CTO. You've got to immerse yourself back into larger, harder problems that have high reward coming out of it. And because of that you've got to develop this ability to you know, sniff out the nonsense from the value.

Ahmad: A lot of people are relying on you to make a decision. I'll give you an example. We can start doing. Just relying on a lot of sensors [00:44:00] that are provided by vendors for our bots to move within the structure. That's great to start with just to get ground through data. But then you become held hostage to that vendor's ability to provide sensors for your bots to move.

Ahmad: That's one approach to do it. And if you keep doing it that way, number one, your competitors have access to the same technology. That's number one. So you're not going to outperform them. A good analogy I use is like trying to build a better taco than Taco Bell by using the same ingredients. If it's the same ingredients, you're going to end up with the same taco.

Ahmad: It's five ingredients at Taco Bell. If we're all using the same ingredients, it's the same taco. So I'm like, no, if you want to build a better thing, you have to go to the ingredients and have, You know, control that. The other thing is you're trying to do something that's a little harder. It's new but you're gonna be the only person with that advantage.

Ahmad: And so it puts you ahead of every other competitor. [00:45:00] But you have to take that leap and your team is looking for you to set that direction for us. It was, Hey, let's not rely on all kinds of sensors. Let's rely on smart people to build rely on computer vision. So now my team makeup is not people who are great at creating relationships with vendors.

Ahmad: No, my team is made up of computer scientists computer vision engineers AI engineers who have cameras that come from a vendor. It can be any camera doesn't matter, 50 or 500 camera, computer vision. And now my functionality of the robot. Is dependent on my team and their ability to build software and features and deployed over the air, not on what kind of vendor relationship I have.

Ahmad: It's a totally different value. Now you've built a company that has value in, you know, the assets are in the people, not in you know, the relationships of vendors and sensors that you really just don't care about. Reselling [00:46:00] at that point, totally different culture. And those people pull in other people that are similar in thinking.

Ahmad: And so when they're faced with a problem, when a piece of sensor or motor doesn't perform just enough depending on the kind of people that you have on the company, they'll either say, okay, this is not doable. There's no vendor that provides this functionality. I can't get more torque on the motor.

Ahmad: So. That's it. We're done. Or you can have a team that could say, you know what? I can wire my own motor. I can build my own, you know, software that actually gives us that extra advantage. Those things are not available to your vendor to your competitors. That's what puts you ahead. That's what the difference between an automation company and a tech company.

Ahmad: A tech company will have people in it that go that extra, you know, that extra little distance to solve things that are not easily accessible to [00:47:00] everybody else. And these people don't just happen to be there. You as a CTO, that's your role. Your role is to say this is doable. This is what, you know, you have to actually do your research, of course, and figure out as CTO, you're the one tasked with figuring, is this even possible?

Ahmad: Is it really valuable? You have to inspire everybody in the company and make up the team in a way. Allows the team to think that way. It's a different approach than just saying, Oh, he's a technology from a vendor. We'll just use that. It's not contract management. It's really technology management.

Mehmet: Absolutely. Yeah. And very insightful as well. Ahmed, like I have a, you know, before we ask about like where people can interact with you, something out of curiosity, like you've worked with These giants like this land, you know, you've been like in maybe smaller organization and now you are in a startup.

Mehmet: So when it comes from the CTO perspective [00:48:00] in managing, you know, because at the end of the day, also, you need to be Cautious with your spendings and then, you know, like, prioritizing the technology investment. So h how, you know, this, you know, make a difference. Especially because, you know, I've seen a lot of people tending now to go and work for either they start their own startups or they want to go and be in a startups, but then I heard some feedback.

Mehmet: They live a shock. Oh. Like, there, like we, we didn't have all these constraints and here like, we need to be like more cautious. So what experience you can share with the audience about this shift from bigger organizations to, to start up on, I mean, maybe smaller organizations.

Ahmad: There are pros and cons to both.

Ahmad: And, you know, you have to shift gears essentially. You have to know in this environment, here's, here are the cards and here's how I play these cards. There's a danger in being in a large organization because it feels like you have unlimited resources. But at the end of the [00:49:00] day, you have to create more value than you're taking out of the company.

Ahmad: Otherwise, the math doesn't work. So with that in mind, you have to think that way within an organization a large one. And at a small organization you're trying, it's almost like diving. You dive and you have you know, enough oxygen in your lungs. You have to come up for air at some point.

Ahmad: You have to breathe again. Very soon. That means this is your funding. If you have enough funding and you can reach the next area where you can breathe great. If not, then you're not going to make it. And so those are 2 different ways to think, but there are advantages and disadvantages to both at a large company.

Ahmad: You have more resources. The danger there is that you forget that it's still a business and you think unlimited resources. So you're not. You don't have urgency. You're not innovating enough. You're not hungry enough. And that is the danger. You have access to more people. So maybe like if you're aware [00:50:00] of the dangers there.

Ahmad: You can do great at a large company. And I advise people generally to start off you know, early in their career as much as possible to go into a large company because you're exposed to a lot more things. But just be aware of, you know, the pitfalls there. And it's almost like starting your own restaurant.

Ahmad: Just not knowing anything about cooking and you start a restaurant like first go work at a restaurant. Learn how to make food. How does the material come in and out? Learn the ropes that you can do in an established organization. And then if you have ambition, go ahead and create your own. But you have that, you know, insight you're not coming in blind now at a startup, the nice thing about startups is that they expose problems early and quickly especially if they're not overly funded.

Ahmad: So it's like a stream. I this is something I learned at Stanford. It was a great analogy. They gave me. It's like a ship in a river and there's a lot of water, a lot of [00:51:00] funding. If there are rocks at the bottom, the ship will never touch the rocks. Everything's fine. But when the funding drops, you're gonna start hitting the bottom.

Ahmad: That's where all the problems are. And so a startup that has just the right amount of funding will still sail. We'll keep going, but it'll hit the problems that are under underlaying at the bottom and they better smooth it out. Really it's a nice balance and I think it's a disservice for a company to Rely too much on funding, especially early in the stage because you will hire the wrong people, maybe for the mission, you will not be ambitious.

Ahmad: You will not have a time constraint that pushes you to innovate and you'll take safe approaches to everything. So it's safer just to go with the most expensive vendor thing as opposed to, hey, I'm Can we build our own? Can we figure out a way, because we don't have the resources for compute. So I'm going [00:52:00] to constrain my computer into a cheaper board, for example.

Ahmad: And now you've got people's brains moving. Now you've built a very dynamic, energetic problem solving team. It's really exciting to be at a startup, but I won't discount being at a large company because that also has its own features. So really it's a mixture of both. I hope that helps. I was like, I'm going through all my experiences.

Ahmad: It's been great.

Mehmet: No, absolutely. I think it makes sense. And then, you know, like even when we talk from business perspective, like this is what usually we advise people, like if you want to start, right. So you need to understand how sales. marketing operations and all these stuff works. So it's better that you spend some time in a large organization.

Mehmet: You see how they do it. Actually, you might even learn from some of the bottlenecks that they have. So you might avoid it later in your startup and you know, you avoid the You know, repeating these same mistakes. I love the analogy that you gave about the ship and [00:53:00] the river. And you're like, I just imagined.

Mehmet: Yeah, like when they are small and little bit funded, like even if they hit the bottom, they can still manage. They will not like break the ship. Right. But when they are big and they hit a bigger button rock. There will be huge issues over there. Really great example. I really like it.

Mehmet: I gonna keep it in, in for myself for reference. Well, sure. Yeah, absolutely. Ahmad, like, I really enjoyed the conversation. Tell me where people can find more about, you know, you and maybe get in touch and also know more about Mytra. All right. I'm on LinkedIn.

Ahmad: My name Ahmed Beman. Mytra is M-Y-T-R-A.

Ahmad: That's really what we're trying to do is have an AI manage all your warehouse. So we're an AI company who happens to build robots. Info at Mytra also you know, you can reach me and the team over there.

Mehmet: Yeah. Absolutely. So for the audience, you don't have to, you know, memorize anything.

Mehmet: Again, I make your life easy, putting all the links [00:54:00] that Ahmed just mentioned in the show notes. If you're listening on your favorite podcast platform and in the description on YouTube. Ahmed, again, thank you very much. I know how busy can things be, especially for a founder like yourself and you know, The fantastic things I advise everyone, by the way, you have, even if you're not in this industry, but you love technology, I advise you to go and have a look at the website.

Mehmet: And even you know, Ahmed, I saw that you have built also kind of a augmented reality thing so you can See the things how they are there. So I appreciate, you know, anyone who push, you know, things forward, like what you're doing, Ahmed. I really, again, thank you for being with me here.

Mehmet: And this is how usually I end my podcast. This is for the audience. So guys, if you just discovered this podcast by luck, thank you for passing by. I hope you enjoyed it. If you did so, please give us a thumb up and follow us and share it with your friends and colleagues. And if you are one of the people who keeps coming, Feedback [00:55:00] comments, please keep doing so I really enjoy reading everything you share with me whether it's good or bad No hard feeling of course, but thank you for all the encouragements that you do And as I say always thank you for tuning in.

Mehmet: We'll be again in a new episode very soon. Thank you. Bye. Bye