 Okay, welcome back everyone to theCUBE's coverage here at ReMars, I'm John Furrier, host of theCUBE. It's the event where, it's part of the re-series, ReMars, Reinforce, Reinvent, Mars, Stance, Machine Learning, Automation, Robotics, and Space. And a lot of conversations, it's all about AI, machine learning. This one's about AI and business transformation. We've got Stepan Pushkarov, CTO and CEO of Co-Founder Provectus. Welcome to theCUBE. And Russ Land, e-commerce, retail data engineering lead at PepsiCo, customer story. Gentlemen, thanks for coming on theCUBE. Great to be here, John. Thanks for joining us. I love the practical customer stories because it brings everything to life. This show is about the future, but it's all got all the things we want, we love. Machine learning, robotics, automation. If you're in DevOps, you're in data engineering, this is the world of automation. So what's the relationship? You guys, you're a customer. Tell us about the relationship between you guys. Sure, sure. Provectus as a whole is a professional services firm, premier AWS partner, specialized in machine learning, data DevOps. PepsiCo is our customer, our marquee customer, lovely customer, so happy to jointly present at this re-invent, sorry, re-Mars. Anyway, Russ. I made that mistake earlier, by the way, because re-invents always on the tip of my tongue. And re-Mars is just, I'm not used to it yet, but I'm getting there. Talk about what are you guys working together on? Well, I mean, we work with Provectus in a lot of ways. They really helped us get started within our e-commerce division with AWS, provided a lot of expertise in that regard and just hands-on experience. We were talking before we came on camera, you guys just on another talk, and how it's all future and kind of get back to reality, Earth. Get back to Earth. We're on Earth still. We're not on Mars yet or on the Moon. You know, AI is kind of got a future, but it does give a tell sign to what's coming. Industrial change, full transformation, because cloud does the back office. You've got data centers, now you've got cloud going to the edge with industrial space as the ultimate poster child of edge and automation safety. But at the end of the day, we're still in the real world now. People got to run businesses. And I think having you here is interesting. So I have to ask you, as you look at the technology, you got to see AI everywhere. And the theme here to me that I see is the inflection point driving all this future robotics change. But everyone's been waiting for it, by the way. It's like been on in movies and in novels is the machine learning and AI as the tipping point. This is key. And now you're here integrating AI in your company. Tell us your story. Well, I think that every enterprise is going to need more machine learning, more AI or data science. And that's the journey that we're on right now. And we've come a long way in the past six years, particularly with our e-commerce division. It's a really data-rich environment. So going from brick-and-mortar delivery to restaurants, vending machines and stuff, it's a whole different world when people are ordering on Amazon every couple of minutes or seconds even, our products. But they'd be able to track all that. Can you scope the problem statement and the opportunity? Because if I just kind of just, again, I'm not, you're in your company, you're in the weeds, you're at the data, and you're everything. But it just seems to me the world's now more integration, more different data sources, you've got suppliers, they have their different IT back ends, some are in the cloud, some are not in the cloud. This is like a hard problem when you want to bring data together. I mean, API certainly helped, but can you scope the problem and what we're talking about here? Well, we've got so many different sources of data now. So we used to be relying on a couple of aggregators who would pull all this data for us and hand us an aggregated view of things. But now we're able to partner with different retailers and get detailed granular information about transactions, orders, and it's just changed the game, it changed the landscape from just like getting a rough view to seeing the nuts and bolts and all the moving parts. Yeah, and you're seeing data engineering much more tied into cloud scale, then you get the data scientists more of the democratization application enablement. So I got to ask, how did you guys connect? What was the problem statement? How did you guys, did you have smoke and fire? You came in and solved the problem? Was it a growth thing? How did you guys connect as a customer with it? Yeah, I cannot elaborate on that. So we were in the very beginning of that journey when there was just a few people in this new startup, let's call it startup within PepsiCo, calling it's not only e-commerce, it was a huge belief from the top management that it's going to bring tremendous value to the enterprise. So there was no single use case, hey, do this and you're going to get that. So it's a huge belief that e-commerce is the future. Some industry trends like from brand centric to consumer centric. So brand product center to Amazon has the mission to build the most customer centric company. And I believe that success, it gets a lot of enterprises are being influenced by that success. So I remember that time PepsiCo had a huge belief, we started building just from scratch, figuring out what is the business need, what are the business use cases. We have not started with the IT, we have not started with this very complicated migrations, modernizations. So can you see the paper? Yeah, from scratch. And so you've got the green light. And the leadership through the holy water on that and say, hey, we'll do this. That's exactly what happened. It was from the top down, the CEO kind of set aside the e-commerce vision as kind of being able to rapidly evolving business place like e-commerce, it's a growing field. Not everybody's figured it out yet. But to be able to change quickly, right? The business needs to change quickly, the technology needs to change quickly. And that's what we're doing here. So this is interesting, a lot of companies don't have that actually luxury. I mean, it's still more fun because the tools are available now that all the hyperscales built on their own. I mean, back on the day, 10 years ago, they had to build it all Facebook. You know, I had people on here from Pinterest and other companies, they had to build all of them scratch. Now, clouds here. So how did you guys do this? What was the playbook? Take us through the AI, because it sounds like the AI is core, you know, belief principle of the whole entire system. What did you guys do? Take me through the journey there. Yeah, beyond management decisions, strategic decisions that has been made, have a separate startup, whatever. So some practical, tactical. So it may sound like a cliche, but it's a huge thing because I work with many enterprises and this like center of excellence that does a nice technology stuff and then looks for the budget on the different business units. It just doesn't go anywhere. It could take you forever to modernize. We call that the Game of Thrones environment. Yes. Yeah, nothing ever gets done. It blows up in the air. Here, these guys, and I have to admit, it's just, I don't want to steal their thunder, I just want to emphasize it as an external person. These guys just made it so differently. They even physically sat in a different office and we were cow-working and built that business from scratch. That's what Andy Jackson talked about two years ago and if you look at some of the big successes on AWS, Capital One, all the big Goldman Sachs, the leadership, real commitment, not like BS, like total commitment says go. And put enough rope to give you some room, right? Yeah, I think that's the thing is, there was always an IT presence by overseeing what we were doing with an e-commerce, but we had a lot of freedoms to make design choices, technology choices, and really accelerate the business. Focus on those use cases where we could make a big impact with the technology. Okay, take me through the stages of the AI transformation. What are some of the use cases and specific tactics you guys executed on? Well, I think that supply chain, which I think is a hot topic right now, but that was one use case where we're using data real-time, sort of real-time data to inform our sales projections and delivery logistics. But also our marketing return on investment, I feel like that was a really interesting and complex problem to solve using machine learning because there's so much data that we needed to process in terms of countries, territories, products, like where do you spend your limited marketing budget when you have so many choices? And using machine learning boiled that all down to, you know, this is the optimal choice right now. What were some of the challenges and how did you overcome them in the early days as it gets being set up? Because it takes a lot of energy to get it going to get the model. What were some of the challenges and how did you overcome them? Well, I think some of it was expertise, right? Like having a partner like Perfectus and Stepan really helped because they could guide us. Stepan could guide us, give his expertise and what he knows in terms of what he's seen to our budding and growing business. And what were the things that you guys saw that you contributed on? And was anything new that you had to do together? Yeah, so yeah, first of all, just a very practical tip. Yes, start with the use cases. Clearly talk to the business and say, hey, these are the list of the use cases that prioritize them. So not with IT, not with technology, not with the migration thing. Don't touch anything, legacy systems. Second, get data in. So you may have your legacy systems or some other third-party systems that you work with. There's no AI without data. Get all the pipelines, get data in. Bootstrap, quickly bootstrap the data lake, lake house, put all the pipelines, all the governance in place. And yeah, literally took us three months to get up and running. And we started delivering first analytical reports. It's just to have something back to business and keep going. By the way, that's huge speed. I mean, this is speed. You go back and head that baggage of IT and the old antiquated systems, you'd be dragging probably months, right? Years, years. Imagine you should migrate SAP to the cloud first. No, you don't need to do that. Just get data, I need data. Stream that data. All right, so where are we now? When did you guys start? I want to get just going to timeline my head because I've heard three months. Where are we now? You guys threw it now. You have impact, you have results? Yeah, I mean that for a marketing ROI engine, we've built it and it's developed within e-commerce, but we've started to spread it throughout the organization now. So it's not just about the digital and the e-commerce space. We're deploying it to regionally, to Europe, to Latin America, other divisions within PepsiCo, and it's just growing exponentially. You have scale to it right now. Yeah, well that's- How far are you in now? How many years, months, days? Well, e-commerce, the division was created six years ago, which is, so we've had some time to develop this, our machine learning capabilities, and this use case particular, but it's increasingly relevant, and expansion is happening as we speak. What are you most proud of? You look back at the impact. What are you most proud of? I think the relationship we've built with the people who use our technology, just seeing the impact is what makes me proud. Can you give an example without believing any confidential information? Yeah, yeah. I mean, there was an example from my talk about, I was approached recently by our sales team, they're having difficulty with supply chain, monitoring our fill rate of our top brands with these retailers, and they come with me, they have this problem, they're like, how do we solve it? So, we work together to find a data source, integrate, just start getting that data in the hands of people who can use it within days, and not talking like a long time. Bring that data into our data warehouse, and then surface the data into a tool they can use, within a matter of a week or two. I mean, the transformation is just incredible. In fact, we were talking on theCUBE earlier today around data warehouses in the cloud, data meshes of different pros and cons, and the theme that came out of that conversation was data is a product now. Yes, yes. And what you're kind of describing is, just give me the product, or find it, and bring it in with everything else, and there's some clean room stuff that people do if they have issues with that, but if not, just bring it in, right? It's a product. Well, especially with the data exchanges now, AWS has a data exchange, and this I think is the future of data and what's possible with data, because you don't have to start from, okay, I've got this Excel file somebody's been working with on their desktop, this is a, somebody's taken that file, put it into a warehouse or a data model, and then they can share it with you. So are you happy with these guys? Absolutely, yeah. Yeah, I should tell the story. What was the biggest impact that they did? Was it partnering, was it writing code, bringing development in, counseling, all of the above, managed services? What was the biggest? I think the biggest impact was the idea, like being able to bring ideas to the table and not just ask us what we want, right? Like, I think for Vectis is a true partner and was able to share that sort of expertise with us. You know, Andy Jackson, whenever I interview him in the queue, he's now in charge of all Amazon, but when he was at AWS, I showed him, he always had to use certain learnings, get the learnings out. What was the learnings you look back now and say, hey, those were tough times, we overcome them, we stopped, we started, we iterated, we kept moving forward. What was the big learning as you look back some of the key success points, maybe some failures that you overcome? What was the big learnings that you could share with folks out there now that are in the same situation where they're saying, hey, I'd rather start Scratch and do a reset. Yeah, so with that in particular, yes, we started this sort of startup within the enterprise, but now we've got to integrate, right? It's been six years and e-commerce is now sharing our data with the rest of the organization. How do we do that, right? There's an enterprise solution and we've got this scrappy, I mean, not scrappy anymore, but we've got our own way of doing data sharing. You kind of pushed through. I mean, you were kind of given charter. It's a startup within the big company, I mean. But our data platform now is robust and it's one of the best I've seen. But how do we now get those systems to talk? And I think Provectus has came to us with, here, there's this idea called Datamesh where you can have these two independent platforms but share the data in a centralized way. And it's... So you guys are obviously have a Datamesh in place, big part of the architecture? It is in progress. It is, we know the next step. So we know the next step, we know the next two steps, what we're going to do, what we need to do to make it really, to have that common method data layer between different data products within the organization, different blockations, different business units so they can start talking to each other through the data and have specific SLAs on the data. And yeah. It's smart because I think one of the things that people, I think, I'd love to get your reaction to this is that we've been telling this story for many, many years. You have horizontally scalable cloud and vertically specialized domain solutions. You need machine learning that's smart but you need a lot of data to help it. And that's not a new architecture. It's a data planes, control plane. But now everyone goes, okay, let's do silos. And they forget the scale side and then they go, wait a minute, you know, I'm not going to share it. And so you have this new debate of, and I want to own my own data. So the data layer becomes an interesting conversation. Yeah, it's like a method data layer. Yeah, so how do you guys see that? Because this becomes a super important kind of decision point, architecturally. I mean, my take is that there has to be some, there will always be domains, right everyone? Like there's only so much that you can find commonality across like in the industry, for example. But there will always be a data owner and kind of like what happened with REST and APIs, how that enabled microservices within applications and being sharing in a standardized way. I think something like that has to happen on the in the data space. So it's not a monolithic data warehouse. You know, the other thing I want to ask you guys, both of you don't mind commenting while I got you here because you're both experts. We just did a showcase on data programmability, kind of a radical idea, but like data as code, we called it. And so if data is a product and you're acting on, you got an architecture and system set up, you've got to have my code, it's programmable. You need your coding with data. Data becomes like a part of the development process. What do you guys think of when you hear data as code and data being programmable? What's that? It's interesting. So yeah, first of all, I think Ross can elaborate on that. Data engineering is also software engineering. Machine learning engineering is software. It's at the end of the day, it's all product. So we can use different terms and buzzwords for that, but this is what we have at the end of the day. So having the data, I will use another buzzword, but in terms of that headless architecture, when you have a nice SDK, nice API, but you can manipulate with the data as your programming object to build, reach applications for your users and give it, and share not as just a table in Redshift or a bunch of CSV files in S3 bucket, but share it as a programmable thing that you can work with. Data as code. Yeah, this is a- Infrastructure as code was a revolution for DevOps, but it's not AI ops, so it's something different. It's data engineering, it's programming. Yeah, this is the way to deliver data to your consumers. So there are different ways. You can show it on a dashboard, you can expose it as an API, or you can give it as an object, programmable interface. So now you're set up with a data architecture that's extensible, because that's the goal. You don't want to foreclose. You must think about that, you must keep up at night, what's going to foreclose that benefit? Because there's more coming, all right? Absolutely, there's always more coming, and I think that's why it's important to have that robust data platform to work from. And as Stepan mentioned, I'm a big believer in data engineering as software engineering. It's not completely separate. You have to follow the best practice as software engineers practice, and really think about maintainability and scalability. We were riffing about how Cloud had the SRE managing all those servers, one person. Data engineering has a many to one-to-many relationship too. You got a lot going on. It's not on managing a database. It's millions of data points and data opportunities. So gentlemen, thanks for coming out on theCUBE. I really appreciate it. And thanks for telling the story of Pepsi. And great conversation. Congratulations on this great customer. And thanks for coming on theCUBE. Thank you, thanks Russ. Would you like to wrap it up with the pantry shops story? Oh, yeah. I think it will just be a super relevant evidence of the geode and speed and some real world outcomes. Let's go, let's go. Close us out. So when the pandemic happened and there were lockdowns everywhere, people started buying things online. And we noticed this and got a challenge from our direct-to-consumer team saying, look, we need a storefront to be able to sell to our consumers. And we've got 30 days to do it. We need to do that work fast. And so we built not just a website, but like everything that's behind it, the logistics, the supply chain aspects, the data platform. And we didn't just build one, we built two. We got pantryshop.com and snacks.com. And within 30 days. Good to meet. The main broker was happy on that one. We'll continue this story. Yeah, so I feel like that, the agility that's required for that kind of thing and the planning to be able to scale from, just an idea to something that people can use every day. And that's, I think, what you're talking about. And that's a great point too, that shows, if you're in the cloud, you're doing the work, you're prepared for anything. The pandemic was the true test for who was ready because it was unforeseen. For us, Majora was just like Eric comes and people who were in the cloud had that set up, could move quickly the ones that couldn't. Exactly, we know what happened. And I would like to echo this. So they have built, not just a website, they have built the whole business line would then launch that successfully to production that includes sales, marketing, supply chain, e-commerce side within 30 days. And that's just a role model that could be used by other enterprises. And it was not possible without, first of all, right culture. And second, without cloud, Amazon, elasticity, and all the tools that we have in place. Well, the right architecture allows for scale, that's the whole, I mean, you did everything right at the architecture. That's scale, I mean, you're scaling. And we empower our individuals to make those choices, right? We're not like super bureaucratic where every decision has to be approved by the manager or the manager's manager. The engineers have the power to make good decisions and that's how we move fast. That's exactly the future right there. And this is what it's all about. Reliability, scale, agility, the ability to react and have applications roll out on top of it without long time frames. Congratulations, that's what I'm on theCUBE, appreciate it. All right, John. Thank you. Okay, you're watching theCUBE here at Remars 2020. I'm John Furrick. Stay tuned, we've got more coverage coming after this short break.