Elizabeth Samara-Rubio

Rahul Abhyankar [00:00] Elizabeth, welcome to Product Leaders Journey. Such a pleasure to have you on the show.

Elizabeth Samara-Rubio [00:05] Thank you very much, Rahul. It's a pleasure to be here.

Rahul Abhyankar [00:08] I want to start with your role at Amazon, AWS, and I'm going to just look at my LinkedIn window here as I read your title. A global head of language, vision, industrial, applied AI and gen AI, use case, go-to-market and business development. So that title does not fit on one line on LinkedIn, let alone on a business card if you had one. Just curious to know why such a long title?

Elizabeth Samara-Rubio [00:35] The role at AWS was a senior manager role with global role leading all of the go-to-market specialists at the worldwide specialist organization for AI services. The scope of the role included having a variety of modalities in terms of AI capabilities, language, industrial, vision, and so forth. And sometimes what modality we're working with is not always clear with the title. So that's why the long title.

Rahul Abhyankar [01:05] So you were working with a lot of modalities with respect to AI back at Amazon and even before that. When gen AI burst onto the scene a couple of years back, I'm just curious to understand what conversations did you have with customers and what were customers most interested in learning and knowing about?

Elizabeth Samara-Rubio [01:25] Excellent. So if you go back to spring of 2023, when everything was really taking off, people were eager to have conversations, and not just the development side, but the executive level wanted to have more and more of these conversations of what is this capability? Why gen AI and what do you think about it? That really was the tenor. What does this mean to my business? In other words, what are you seeing as potential applications, use cases where this new capability can begin to maybe reshape how I'm setting up either my outbound support services or my internal product development capabilities. But it was really centered about, help me find a compass for how to navigate this new capability that looks great. I just don't know really how to apply it.

Rahul Abhyankar [02:12] And did you have a mental model of how to answer those questions?

Elizabeth Samara-Rubio [02:17] There's two parts to that answer. There's the personal and then there's obviously the company. From a personal perspective, the mental model was really first human. Why were people taking off so quickly to try to get their arms around this technology? What was driving that interest or a sense of urgency to try to understand it? And obviously by the summer of 2023, a sense of urgency to try it. My mental model was more about how do I understand the business drivers, the personal drivers that were behind the meetings that we were having. And ultimately, how do I frame where we are today with AI and how do we begin to exercise some of these new generative AI capabilities that we're now talking about? Because in my role at AWS, I was already supporting customers with my team around global deployments, around AI services. So there was a lot of questions about how do I go from here to there? And when customers grapple with some innovative new technologies, it ranges from a sense of curiosity to a sense of urgency. So that aspect of early adopters, innovators versus people in the mainstream part of the bell curve looking for proof points of that technology having found its legs before they adopt or do something with it.

Rahul Abhyankar [03:30] So did those conversations give you a sense of how to evaluate where customers were on that spectrum from curiosity to urgency?

Elizabeth Samara-Rubio [03:38] Absolutely. That's a good way to say it. The number one thing was to understand where is the customer in terms of their maturity around digital transformation. That gave me a reference point for balancing how much of the conversation was going to be more about curiosity versus a sense of urgency to build on something they already have. And the reason I say that's important is because in order to do generative AI, the customer had to have some foundational elements around how they're managing their data. Without that, before they can exercise the generative AI capabilities as part of their processes, they were going to have to maybe take a step back and do another level of investment around setting up the infrastructure just to be able to capture the data that they needed in order to create these key applications. So it really boils down to getting a sense for where is the customer in the maturity cycle so that I can better frame how we're going to have a conversation about what is possible.

Rahul Abhyankar [04:40] And there is an element of Amazon being very famous for working backwards and that methodology that applies to building products. But that working backwards methodology and that mindset, did that also apply to go to market, business development? How far reaching is that methodology within Amazon?

Elizabeth Samara-Rubio [05:00] Absolutely the fabric of how we think. I say "we," it's still part of me, thinking backwards from customer outcomes. The best way to have a conversation with a customer is to understand where they want to end up. Because it will be difficult if the customer is leaning in with a sense of urgency around this technology, again, 18 months ago, almost, they were trying to solve for something. The question is, what were they solving for? And then we can work backwards from there. The go-to-market, it was really thinking through all the outcomes we were understanding and learning. What was the pattern we were beginning to see as to where this could go and where it had some legs and where perhaps would remain a little bit more exploratory for a while. And working backwards from that gave us a roadmap for how to temper, how do we begin to look through the different services that we can provide? And then ultimately, how do we go to market, not just alone, but obviously with partners?

Rahul Abhyankar [06:00] How did you actually, for listeners to relate to what does working backwards mean? And in the context of go-to-market, working with partners, customers, how does that actually come to life?

Elizabeth Samara-Rubio [06:15] Two examples. And these are examples from a year ago almost. So the generative AI was at a different stage than it is today. But the two examples that jump out to me—I was talking with the CEO and COO of a biopharma company. Trying to say, okay, maybe this capability can help us do a more efficient and therefore faster process around being able to do some discovery. And in listening to the conversation, of course, we were talking about RAG already, making sure that we had the data already prepared. And of course, what were some of the KPIs they were looking for in order to assess whether this was the right investment for them. One of the things that I recall, just understanding how they were doing discovery today, one of the things I shared with the CEO is that with all the work that you're going to do, fine tuning and part of that using prompt engineering, you may want to consider how your prompts are going to become your form of IP. Because you have to think through, the prompts coming from your scientists are the mechanisms they're already using as a basis for discovery. So now when you convert that into a prompt mechanism, you may find yourselves where those prompts become a form of IP as part of your development process. And that was different. She was like, I never thought about that. That's one of the things you want to take care of is how you're doing the prompts as part of discovery because that is your IP.

Rahul Abhyankar [07:45] That's a fascinating perspective. I've not heard this before, is the prompts can be thought of as IP because that really leads to how that research and that discovery is happening specific to that domain.

Elizabeth Samara-Rubio [07:58] Exactly. And in that same conversation, that's when I began to also realize that these companies were not just looking for efficiency gains or perhaps innovation in the form of a better discovery outcome, but they were also beginning to explore that perhaps with the generative AI capabilities, what they had held before as pretty close to the chest, their process could become a new business model of what they do. That was interesting because then I'm like, so enterprise customers are going to become ISVs if you think about it. They could become that. That was eye-opening.

One of the last conversations I had with customers around generative AI, it was a company here in California. It's a room for like 24, 26 people in a small conference room. We were talking about what is generative AI? What are some of the use cases we're beginning to see? A third way into the conversation, the CIO asked, or actually stated, I don't need help with thinking through use cases. I have about a hundred of them. Can you give me a template for how to make an investment business case for generative AI? She goes, that's what I need. And that was very powerful because it was the first time someone started talking about the business case and investment versus just a curiosity and exploring what is possible.

Rahul Abhyankar [09:15] So that clearly is a good signal to understand where on that spectrum from curiosity to urgency they are. So a couple of thoughts that you sparked there. One is you mentioned enterprise customers have the opportunity to become service providers. So can you elaborate, explain that further?

Elizabeth Samara-Rubio [09:35] We both know that in generative AI, it was quickly identified that the value of the application developed goes up considerably when it's vertical specific. And if you think about the way when you look at verticals, in order to train these models, the amount of data you have by vertical versus across verticals is very different. And the third element is the data that is vertical is probably higher quality. So customers started looking at their data very differently and for their vertical domains. And they were trying to say, well, maybe what I've been using as an internal process to process my data for product development purposes, that could become an extension of not just an internal process, but something we can provide externally. One example of that was—I can't say the customer name—but the customer's business model is to source peer-reviewed papers and so forth. And obviously to do that, you have to have a lot of mechanisms and access to information to do that. But that type of data access gives someone the opportunity to be able to turn that around and say, well, maybe we can develop an application of our own and create a business model around the access or the searchability or the development of new data as part of our service versus just being one of consumption.

Rahul Abhyankar [10:55] That's fascinating. It will be interesting to see how enterprises take advantage of that and really come up with those really innovative business models.

Elizabeth Samara-Rubio [11:05] I think it's going to happen.

Rahul Abhyankar [11:08] And then the other thing that you mentioned, which I thought was really interesting was the investment case. So a gen AI investment case, a template, what does that look like? How different is it from the business cases that a lot of people have been doing pre-gen AI?

Elizabeth Samara-Rubio [11:25] My perspective, not reflecting obviously my previous role: in the past, it was largely around, I would say almost commoditized horizontal capabilities around customer support. Generative AI became yet another method or mechanism for trying to improve the quality of the service delivered through customer support functions. So could you improve your throughput at a number of calls completed in a period of time by one person? Or can you improve your close rates of a service ticket in a few number of people? So that was a level of the business case. I would say almost for the first chapter of generative AI, it was, can I improve what I already have today? That business case is leaning more towards automation.

What we're seeing some customers explore with is, if I can now have multimodal capabilities at the edge, that gets me one step closer to be able to do a few more things with closed-loop automation. And I'm thinking more of the industrial sector. It's not just what you see coming down the production line, whether that's the product or the equipment of the process, but it also becomes what you can see, hear, or measure in terms of temperature or voltage. When we introduce all of these senses onto an operation where you have heavy equipment in production mode, the generative AI use case or investment case becomes one of, now we have a 24/7 system that can hear, listen, read, and most notably adapt itself to the actual process. So now we're talking about increasing throughput, increasing yield, and potentially creating higher value-added workloads for the teams we already have in place.

Rahul Abhyankar [13:05] So let me see if I understood that. The first example of the contact center and customer support, that's where generative AI and that investment case is really about enhancing the existing KPIs that that contact center is measured by. But then in this industrial use case, that's about creating net new capabilities or having a virtual or an AI operator on the industrial floor that's effectively monitoring the processes and the systems and the machinery that's working and the way it's working and then learn from that.

Elizabeth Samara-Rubio [13:35] That's correct. And I like the way you said it, Rahul, AI operator. It becomes that next level of augmentation. It's foundational. I think we both know that it's been a while that the industrial sector has been trying to accomplish a closed-loop automation platform.

Rahul Abhyankar [13:55] Fascinating. So Elizabeth, you started at HP as a product manager. You were product management for the support and services division. How did you wrap your mind around product management and applying that to support and services?

Elizabeth Samara-Rubio [14:10] The first thing I have to smile about HP is my first few months at HP, I started in their sales office in Austin, Texas. Got lots of visibility as to what it is to support the customer and the sales teams, because that's foundational to now moving into a North American role and then ultimately into a global role at HP. From sales to channel marketing and then to product management. By that time, I was already understanding that there's a lot more to success of adopting the hardware systems that was based on service, the level of support and services that we delivered upfront or as part of after installation. So for me to get my head wrapped around it, was understanding what it took for deploying the systems, installing the systems and maintaining the systems, upgrading the systems or adding to the systems.

About that time, it's the late 1990s, early 2000s. So there was a huge push to create more data centers. If you can remember what we were doing then. And they said, okay, you want to come here and do market research on what's it going to take to do these types of services? You can come in at nighttime and watch us how we take everything down. All these servers have to be brought down, packaged, and moved to a different site where it's going to be bigger and we're going to add more. They would literally say, get here by around 10:30 PM, because that's about the time we start. Those types of experiences, Rahul—one thing is to read about what it takes to do something. The other thing is to actually go see it. That was quite insightful of what it takes to actually provide a level of support or a type of service that helps those types of relocation activities actually happen flawlessly so that by the morning, the systems are back up and running in a very different location and back to business, so to speak. Nothing beats getting out there in the field and actually seeing it.

Rahul Abhyankar [16:00] Absolutely. We had a saying when I was working at McAfee, which is nothing important happens inside the office.

Elizabeth Samara-Rubio [16:08] Exactly. I consider customer contacts oxygen. Without that, there's too much carbon dioxide.

Rahul Abhyankar [16:15] That's great. But then, when you think about servers as the product, after the installation and the support and the services that it requires to keep that system running, so you had that product, if you will, as you were responsible for. And so that's very different from just building the software and building the product itself. So did you have to convince the product team to actually take some of the learnings that you had and incorporate that into the product to make them resilient or make them reliable, available, and so on? So just curious, how did that happen?

Elizabeth Samara-Rubio [16:50] HPE was different back then. I mean that in a very positive way. For us, the best way to give feedback to the product managers was to be able to show basically their service tickets, number of service tickets or severity levels, or even just customer testimonials to try to demonstrate this is what we're dealing with here. Because the spectrum of services that I covered, it was all the way from global installation of systems—we're talking hundreds, if not thousands of systems—integration of those systems and pre-testing before they're shipped to on-prem and then installed, all the way to what was back then called the five nines of uptime. All of that was threaded together with our systems in order to provide that level of feedback back to the product managers. It was the only way to continue to raise the bar, but also there's an expectation from our customers that you're going to get better. This is going to be continuous improvement. That was one.

I'm happy to share also one of the most interesting—there were two really interesting projects that I loved doing at HP. One was when I was approached to become the fifth person on a five-person team to make a business case where we would combine the for-profit installation services with a cost center that was an integration center. It was through that exercise of putting a business together that I learned two things. One, how to pitch for investment. And number two, it was also how to transform a cost center into a for-profit center. That was an inflection point with HP.

And then the other inflection point with HP, again, product management, just thinking about this as a product leader, knowing when you get this rare chance to be the first at something in a big company. At HP, I got this opportunity with great leadership support to actually be the lead for building the business case for the first website slash portal for system support. That was the next level of maturity in being able to deliver our services and start delivering the services in a very different velocity.

Rahul Abhyankar [18:55] And that aspect of taking a cost center and creating a business case and turning that around into a profit center—that's not an experience most people have. So I'm just curious to understand what were your key learnings as you did that?

Elizabeth Samara-Rubio [19:10] Thinking back to the business case, the number one thing was, how do we define the markets today? How does that definition of the markets we go after change if you do absorb this cost center? How does the value proposition change from a customer's point of view when we combine the services? And then ultimately, what does that do to our business model? If we do that, the benefit is I had to do all that work, all of that primary and secondary research to create the framework for how we were going to—how much more of the market can we go after? How are we going to price this? How are we going to define these services? How are we going to roll out these services globally? And the number one thing was to be able to show that the investment of combining these services was going to give us not just more in terms of the top line, but it also opened up doors for far more services down the road.

Rahul Abhyankar [20:05] So while you're doing these business cases, you mentioned just now an opportunity to do something for the first time within a large organization. Did you encounter any things that you had to overcome?

Elizabeth Samara-Rubio [20:18] Yes, many. There was a general manager, first name is Wade. We had already traveled to a couple of countries as part of getting ready for the launch of the new business. He took me out to lunch and he sits me down and asks, Elizabeth, have you ever driven shift, you know, like manual cars? I'm like, once or twice, really bad at it. And he's like, well, you get the concept, right? First gear, second gear, fifth. I'm like, yes, totally get it. He said, okay, Elizabeth, you always seem to be in fifth. You're going really fast and you're getting things done. That's great. Sometimes you have to be the audience of your stakeholders and understand, where are they? What gear are they in? And I didn't understand. I had to ask, where is this going? And he said, well, you want to get things done. You're ambitious. Philosophy is what sets your pace, but you need others to get things done. This isn't going to happen just because you go faster. This is going to happen because you took people with you. And it was that time when he said, you're going to have to learn how to read your stakeholders and know how to shift.

And I was like, well, why? I don't get it. I mean, in your twenties, you're wondering how fast can I go? And I was very naive. I'm being very open with you. Very naive in the sense that my response was not something I believe in anymore. At that time I said, if they can't drive at the highest speed possible, then why drive? And he said, Elizabeth, you're going to need people with you to make the kind of changes we're talking about. The words were enough for me to change, but the words stayed with me as I grew in my career.

And today I call that clock speed. I'm looking for understanding my clock speed. I'm looking for my team's clock speed. I'm looking for my customer's clock speed. And a market. And that's important because at the end of the day, we really can't rush those things that are outside of our control. The best thing we can do is understand where is it, and then if the system wants to go faster, understand where is the governor in this system and how do I help you change your clock speed, assuming you want to.

Rahul Abhyankar [22:30] That's an interesting concept, clock speed. And when you talk about knowing your own clock speed, that's obvious in terms of how fast you want to go, the urgency that you bring to something. But then understanding another person's clock speed or the team's clock speed or the company's clock speed, which in case of HP at that time potentially was not aligned with your clock speed. So what are you looking for when you're trying to understand someone else's clock speed?

Elizabeth Samara-Rubio [22:55] Two things. One, how do they approach decisions? What information are you looking for? How do you define the problem you're trying to solve for? And then what are you doing to get your answers? I can learn a lot about a person's clock speed just by the way they are going about making a decision. It's not a question of good or bad. It's understanding how they're processing information. How are they approaching the ways they can solve for that situation? How are they framing success in that situation? Then I know I can sense your clock speed.

Big picture, what do you want to get done? I want this to get done. Well, how do you know we got there? Okay, it's a metric. It's either calendar, a time, deliverable or something. Great. What are the top two things you need to get that? And then as soon as I hear that, if I'm not getting a clarity of thought that they've understood the one or two things or big rocks that you need to get your arms around in order to get there, then I'll help that person. I'll do a little scaffolding and then I say, okay, so now that we identified the two big rocks, how are you going to understand what the nature of those rocks are, what information do you need? And if a person starts picking it up, then I know that, okay, now I got the flywheel going, they're gonna go. But if it's taking a while, I'm gonna realize, this is a different clock speed than mine, and that's okay. But I'm gonna have to adapt. At that point, I can tell if the person and I are at different clock speeds, and that's okay, I will adapt.

Rahul Abhyankar [24:25] And so I guess what I'm getting at is not just a question of somebody being a busybody and having a lot of just activity around them, but it's really a question of how clear is their thought process? How deeply have they thought about something which then gives them the propellant to move faster?

Elizabeth Samara-Rubio [24:45] That's right. You said that so well. It's absolutely about the clarity of thought. And it's not just where you're going to go, but what are my one or two things I have to unblock and how am I going to unblock that, and having plan A, plan B, plan C, and thinking that very through, and then the rest is going to be faster.

Rahul Abhyankar [25:05] And I love what you said about when you encounter someone who's not at the same clock speed as you, you said that it's you who have to adapt to their clock speed. Which is fascinating because when you as a leader understand the different clock speeds of different people on the team, then you have a little bit of a benchmark to say, okay, how much do I have to put in to bring this person along or at least have them come up to your clock speed in that clarity of thought.

Elizabeth Samara-Rubio [25:35] Yes, Rahul. And I think this goes back to the human element. This is not about, like you said, it's definitely not about busybody. It's back to the clarity of thought. It's very important for me as a leader to make sure the person knows that I am supporting them to be successful. At this point in my career, we'll hit goals. But did I help somebody or did I help the team get to where they wanted to be personally? It's important to have the team together because we're going to definitely change things. But we're going to have to go through so many walls together that we have to go in together. We have to come out together.

Rahul Abhyankar [26:15] So then you were at Accenture, which is a completely different sort of a role, and Industry X.0. What does that mean, Industry X.0?

Elizabeth Samara-Rubio [26:25] Today, it's a very big business for Accenture. At that time, when I first came, maybe it was about 15 months old. The idea was redefining industries through digital capabilities. That meant at that time, obviously cloud is one of them, but also embedding automated capabilities, whether it was smart products, consumer appliances, coffee makers. Engineering side, would be at that time, it was the beginning of what can we do with digital twins? And then there was a third to Industry X, which was digital manufacturing and operations. So Industry X was about redefining these three pillars by embedding more digital capabilities across these practices or functional areas for companies.

The other area of growth was developing new programs. And that was around—I started focusing on an extension of manufacturing and operations, bringing in AI capabilities with vision systems. And obviously that's a thread that had started long before I joined Accenture, which is why I could do it. And last but not least, was the strategic planning around Industry X. So those elements of understanding the maturity of an industry, strategic planning around that—and when product people look at markets and industries, they tend to look at the attractiveness of a market.

Rahul Abhyankar [27:50] So how do you take, or I guess what you learned at Accenture across Industry X and doing that, how do you think that can be translated to applying towards market attractiveness, industry attractiveness, or the competitive dynamics within an industry that product people need to think about?

Elizabeth Samara-Rubio [28:10] Good question. The first thing for significant growth in a market is know how it's defined today and start figuring out how to redefine it. Because if all we're going to say is, it's an attractive market because look how big it is and look, it's just a few players in it, then it's probably going to be an incremental me-too play with a little differentiation. And it's going to be a single-digit year-by-year growth thing. So the first thing I learned was, step one, define it today like everybody else does, and then step two, figure out how to redefine it. Once I redefine it, then I can change it. And then when I change it, I have to validate it.

If you think of traditional exercises in market attractiveness, think of the Mekko charts that people do. I don't know if you've ever done those for strategic planning. You are basically mapping an industry by market segments and you're looking at how large the market is versus its growth. What I would do is to cross the bar for investment, I said, if we just take this, we have to change the way this market map is laid out before we can even make a pitch for investment. Otherwise, it's not going to have the returns that we needed for the kind of investment we're looking for, which is for growth. So when we did that at Accenture, it was really trying to forecast what our company had all been trying to do, but hadn't been able to accomplish yet. And a lot of that was back to our opening conversation. It's been a while that there's automation on the table. Now that we were looking at AI systems, more automation became viable. But the only way to make these viable systems or viable AI systems practical was to make sure that we could become the trusted partner for companies to make that investment and let them know you're going to be okay if you go down this path. You're going to be okay if you're going to start using a cloud provider. You're going to be okay when you're starting to connect your data in the cloud with your MES system on-prem. It's going to be okay. And that's what we needed to step into. Hopefully I answered your question, but it was literally a rethinking of markets and applying what we knew was already happening and knowing that we had some gaps in our delivery and we went to fill them.

Rahul Abhyankar [30:20] The question that came to my mind, I think what you said makes perfect sense, giving companies a roadmap of how they can start where they are, leverage technologies, and Accenture and companies like that have the expertise in bringing companies along. But in that sense, in the market, are you working with early adopters or are you working with late adopters, the mainstream market that's the biggest part of the bell curve?

Elizabeth Samara-Rubio [30:45] I smile because all of them are in one company. So I came up with the term: if you want to do the innovative stuff, you're going to have to look for the pockets of possibility. Pockets of possibilities—I'm looking for the person who's been trying to do something that's very different and just hasn't been able to anchor themselves with something to show. Something to say besides a presentation or besides an Excel sheet. Someone who's been trying to demonstrate what is possible. Those are the people that we anchor ourselves with so that we then can cultivate a narrative to the rest of the stakeholders who are perhaps more the laggards, to use your words. And I say it exists everywhere because in large companies, you do have teams that are pockets of possibilities and then you have the other teams whose only job is to sustain today and make sure it works.

Rahul Abhyankar [31:35] Interesting. I like that term, pockets of possibilities. And those are really opportunities to understand what are people in those companies trying to do, trying to create those new business models or new capabilities, but they need some help.

Elizabeth Samara-Rubio [31:50] They do. And that goes back to the human element. When we as product managers, sometimes we don't realize that the person who takes a chance on us, the person who is in that pocket of possibility, their career is on the line. So it's not just, did we meet all your requirements? Did we make you successful? Because they're the ones negotiating risk. They're negotiating the investment case, they're negotiating someone pitching an alternative. And so when they make a commit to you or me as a product manager, it's a commitment at a human level. And I would say, don't say I'm going to go do some amazing, crazy, innovative things with you if you're not ready to step in and also make sure the person's going to be okay.

Rahul Abhyankar [32:35] Love that. So coming to Sima.ai, you are here as Chief Business Officer, all about machine learning at the edge. What does that really mean, machine learning at the edge?

Elizabeth Samara-Rubio [32:48] I'll give you the technical answer in a second, but I'll give you the human level experience first. ML at the edge means there is an appliance, there is a system that is making a prediction in the order of magnitude of 500 to 800 milliseconds. Really fast. And it's doing a prediction that you had decided is important. Technically, you can have that same prediction in the cloud. Nothing wrong with that. But if your system, if your process requires sub-second latency times, you will be running that application at the edge.

So if you think about there's time series data and then there's vision. We currently are focused on enabling customers—first modality is vision. And vision is a huge payload. Sending it from an on-prem appliance to the cloud may take more time than you can afford. And also given that size may also exceed what your allowable investment case may be. So in those cases, we work with customers to drive those inferences around a vision system at the edge. What do they do? So in the industrial sector, they can do anomaly detection. Anomaly detection could be product quality or process automation. And you're looking for things that are trending in the wrong direction outside of acceptable thresholds. You're looking for something that's obviously deviated from what is acceptable. Obviously it's a binary classification, good or bad. And those decisions are literally made within half a second, no more than a second. So we work with customers to do that in the industrial sector.

The other element is embedded AI. So this is the idea of putting an AI system into a piece of equipment for purposes of improving the equipment's performance. So that is a game changer in order to make this application of an AI vision system viable as a value adder to their apparatus. Most customers are training in the cloud and then porting over their models to the edge. That means by very definition, they need to port models effortlessly. So our entire software design is about making sure that effort comes at one, ease of use, but number two, a very low loss of efficiency or accuracy to be able to put that model over to our system.

Rahul Abhyankar [35:00] Interesting. And given that vision has a heavy footprint, like you said, the cost of training models based upon vision data, I guess that cost is significant then.

Elizabeth Samara-Rubio [35:15] It can be. Let's just say an order of magnitude is more affordable than a generative AI application. How's that?

Rahul Abhyankar [35:25] Oh really? Okay, interesting.

Elizabeth Samara-Rubio [35:28] Yes. Because again, where we are with AI is we have established to date that if it's a vertical data set, if it's a task-specific data set, by its very definition, you already are working with a more narrow set of images. So you're not running these huge, capture everything and possibilities. You have that well-defined scope and all of a sudden the training is in about weeks.

Rahul Abhyankar [35:55] Interesting. So the dataset is fairly specific. I imagine the quality of the dataset is also very high.

Elizabeth Samara-Rubio [36:02] It has to be. And I always say with vision, because I've been doing this for 10 years, it's very important that the optical configuration—camera, lensing, lighting, angles, working distance—are vetted out early before we go down the path of taking up a customer's time with capturing the thousands of images we're asking for.

Rahul Abhyankar [36:25] Fascinating. I want to be respectful of time here, but a few questions that I want to cover are—when you talk about clock speed, this aspect about knowing yourself and how fast you were wanting to run at, was this part of growing up, upbringing? How did this come about?

Elizabeth Samara-Rubio [36:45] Thank you, Rahul. You're making me smile because now I have to think about the origins. A little bit about how to get to clock speed growing up. There was always a sense of, you have to go get something. Like if you want something, you have to go get it. A sense of urgency, a bias for action. Know what you want, but you're going to have to go get it. There's no waiting. That was instilled in me through my own personal experiences growing up. I was not born in the US. My family is from Colombia. And so as most immigrant families, you come here for change and change doesn't come easy. And you have to go and hopefully ask really hard questions and hopefully most of them are smart questions early so that you can figure out how am I going to get this done. So that was the genesis, and because that was so much of what I had to do growing up, I just thought everybody did it that way.

Rahul Abhyankar [37:40] So influences from growing up, lessons that you've learned growing up—are there things like that that you've carried over into your professional life and just looking at things in terms of, as leaders, we're all trying to do more with less.

Elizabeth Samara-Rubio [37:55] So, Rahul, my mom, she was a teacher in Colombia. She has stories about getting on a horse to get to her school in the mountains. But when she came here, obviously, like most people who want to teach, they're going to have to go back to school and get their undergrad and get their certification. And she did all that. And she got her master's degree and then she started her PhD. And she always had a high bar for do the right thing. And she would always say, if you're gonna do something, do it right or don't do it at all. I was like, wow, gosh, thanks. So inevitably, that always repeated while I was growing up was something in my mind. I always have a very high bar because I can't, I still hear her voice in my head. Do it right or don't do it at all. So those are one of the things that were foundational to me.

And obviously when you're trying to do a lot in your life, and you just don't have—I mean, we're all still searching for the 25th hour in a day. You have to figure out, big picture, what do we really need here? Big picture, how do we do it right? Big picture, am I the right person? Is this the right thing for me to work on or is this something for someone on my team to work on or maybe it's not even the right thing for us to work on? But that's how to navigate, you're saying, do great with less is know your bar. I bet it's pretty high. And then ask the big questions and see if ultimately you're going to have to make a decision. Are you the right person to do it?

I feel that I've been very lucky. No matter how much work it takes to get something done, it takes a little bit of people caring extra for you or people making time for you. I'm very grateful for that. I'm very grateful.

Rahul Abhyankar [39:35] Beautiful. Thank you so much, Elizabeth. Such a great pleasure, such a great conversation. Thanks for taking the time.

Elizabeth Samara-Rubio [39:42] Thank you for making the time, Rahul. Look forward to it.