Build & Scale Enterprise Grade Self-Checkout - Habeebulla M & Deepak Bajpai

This talk covers the transformative journey at Lowe's, focusing on building and scaling self-checkout systems. The speaker discusses the company's significant digital presence and the need for a more flexible and robust self-checkout solution, which was hindered by existing rigid vendor systems. The approach involved defining key performance indicators related to system availability and flexibility, ultimately leading to a modular, layered architecture designed for better performance and maintenance. The speaker emphasizes foundational principles of engineering to guide the development process. Additionally, the talk touches on the commercialization of Lowe's technology, particularly a point of sale solution that integrates various retail functionality, showcasing how Lowe's aims to contribute to the retail industry with innovative products and services.

So, what we are going to cover here especially in an age of uh AI, I think I don't know how many of you probably would have attended for multiple days now. Uh one thing I can promise you is you'll not hear so much of an AI here. Okay? So, let me start with little bit of I think thank you for that wonderful introduction. But then, a little

bit about Lowe's. For those of you who probably have not heard much about Lowe's, it's the Fortune 100 company uh second biggest home improvement retailer in USA catering to close to 330 million registered customers and then clocking over 16 million transactions uh week over week. So, we have a significant digital presence plus close to 1,800 very large format stores which forms the backbone of our retail system

back in USA. Again, uh the legacy of Lowe's really goes back to almost like 100 plus years now with always an eye for technology. But then, as with any of the company, the biggest transformation really the pivotal moment came to us back in 2018 when our current CEO um and the chairman, Marvin Ellison, joined the company. And he brought along Seemantini Godbole, who is the Chief Information

Officer who leads the technology function globally. And then, we we had um Mr. Ankur Mittal joined us as an MD and the Chief Technology Officer back here in India who leads the GCC. I am part of the GCC. We are close to 5,000 very strong talented associates spread across both business function as well as the technology. Um on various functions including your supply chain marketing finance, digital

stores all over and and I'm kind of proud to say we are one of those very few GCCs I think who have figured out a way to play into the entire value system for the mothership. Um both from business as well as the technology functions. So, when we started this I think the the pivotal moment that I talked about from 2018 onwards there is a lot of

a transformation that we have done in a company but then what I'm going to talk about is one such the story of one such which is building and then scaling the self-checkout systems. So, how many of you have probably used any of the self-checkout systems here in any form or a fashion? Great. I I see quite a few hands here. So, what I thought is we'll probably

structure this into a three sections on a very very broad level, right? So, one uh what was the problem that we are trying to solve? Why did we even do what we did? Second is how did we go about it? And third of course is looking back now connecting the dots what did we learn? Um again not to keep any suspenses here let me answer the first

the last question first. What did we learn? What we learned is the foundational first principle thinking always pays off. And more so in the engineering. So, we'll see how it is. Okay. Uh it's not like we never had a self-checkout. We had a self-checkout in our ecosystem. Um we like we we had close to our 2,000 stores, but then we could afford self-checkout systems for 1/3 of

those because of various reasons. We couldn't scale. So, the I think the best question I thought should we ask is why did it not work for us? Right? Um why did any of such systems fail? So, when you think of a self-checkout, it is one of those unique systems which sits in a very secured environment, we usually in the four walls of the stores, but then it

is for the customers to come and then use it. It is exposed to the customers. So, inherently what that brings up is the notion of security, right? So, you have to you have to be build a system which is very very secure, strong on security. And in the traditional systems, traditional architecture or the design pattern, security always had a unfortunately, trade-off with rigid architecture. And that is

what we had even with our ecosystem as well, very rigid. We had a system which was completely vendor built um right from the hardware to the software, all of that. So, because of the rigidity, your the workflow, the compliance updates, whatever that you have to make a changes, it's all took its own sweet time. Then the high cost. when I talk about self-checkout, I'm not talking about

just the experience, right? So, self-checkout has like three major modules into it if you look at. One is the hardware in itself. You have the printer, scanner, then you have a compute, then you have a pin enabled like your payment devices. All of them have to come together. Then on top of it, there has to be a service layer which is kind of talking to all your

devices to make it look like one. And then comes your experience on top of it, right? So, all three put And when you see that as a combined package, one package, which is given by the vendor. Now, again, there is a nuances of retail systems, right? So, you have something we are into a home improvement. There may be somebody who is actually a big box retailer, somebody

who just sells the grocery. The systems which are built out there in the market are for the generic needs. So, the example I can take is the self-checkout systems that you are running in. Hard a big space for a weighing scale and to put stuffs which are more of your groceries in nature, which are just not needed for our business. So, but still you ended up actually

picking that up. And hence, there is going to be a pretty high cost that you end up paying because of the way this is all kind of structured, right? And and how it is bundled. Needless to say, that goes with limited flexibility. So, you will have because of you will not have a choice to make whether I need this part of a hardware versus that. What would

work for your line of a business? So, you really have to work through all of that. So, you have to work in a constrained environment. Each of this has its own failure points and hence come up with their own downtime. Then, of course, the limited capabilities and difficult to customize. There is no customization at all. I'll tell you a story going back at COVID during the COVID

times. We all know this uh the disinfectant that we all used, right? So, we did it the same thing even in the low stores as well. I don't know whether it killed the germs or not, but then it killed some of our hardwares. Okay? And then one of that is exposed hardware piece of hardware is your PIN entry device, which you use for taking the payments. And

then because of the structure that we had, which is all very, very closely knit and then very dependent on each other, comes as a collective unit. One small component going rogue, now your complete unit is left unusable. And then because thanks to the supply chain delays at that point in time, we had to wait for 3 to 6 months. the outcome of that, the entire system is

not available. So, what did we do? How did we really go about? I think the answer is simple, right? So, we had to solve for everything that was a problem for us, okay? In engineering, there are different ways of looking at things, right? So, one is of course you have to first define what is your key KPIs are. Key KPIs are not just for measuring whether or

not you're successful. This is also one of the very important factor that kind of goes into what do you want to design your system to be? How do you want to design it? What is the level of a control that you want to have so that you will be able to eventually give that deliver on your key KPIs. In our case, our KPI is the availability. The

availability and then of course we have to be available to it has to be affordable, it has to be non-rigid, and then flexible with the hardware, nimbleness with the software, all of this, right? So, so we thought what we said is let's make it more the modular better, right? So, how did we really go about all of this? we said my number one is the flexibility and

hence I have to come up with an architecture which is more layered in nature. Okay. A decision is also to make if your availability is the number one thing, whether or not you want to put something on the edge or you want to put something on the cloud. Of course it's a debatable, but then by and large, usually the availability is high on the edge when you

compare to the cloud. But then what it comes up with is the challenges that you have to do with the distribution, right? How do you distribute? In lowest context, when I say edge, it is 80,000 of those, right? How do we do that? And hence, there is a What team came up with is a layered architecture where we said, "Okay, what is that I need? I need

a first of all, operating system. Then I need a service layer in between. Then I a layer where it kind of provides the runtime. And then comes the application. Bundle all of this together, but then still maintain the isolation between all of them. So that when I have to make a change, I don't end up deploying the entire bundle all the time. Right? On the edge, an

immutable design. When I talked about the rigidness, that is all because of the security. How do you address all of this? You can make your system completely immutable. Design it in a way every time there is a restart, it kind of goes and pulls the fresh copy of it. Right? So that it remains immutable. Security at the core, anyway, it is already built in. The upselling opportunity

is multi-language support, intuitive customer All of this is more around how do you build your experience to be. with a strong UX in mind. I think I'll just In the interest of time, I'll skip on all of these some of these. But then here is here is the thing, guys. So Okay, that slide is missing here. But then what I would say is end of the day,

what really helped us is again going back to the foundational thinking, going back to asking what are we building and then why are we building it? And then based on that, go back to the first principles of engineering and then ask those very, very basic questions as to what is that you want to do and how do you want to structure your uh architecture to be to

deliver on the key KPIs. So with that, I think uh I'll invite Deepak to come and then talk about some of the commercialization pieces that we have done around this. Over to you. Hey. thank you Habib. Uh my name is Deepak Bajpai and good afternoon folks. Folks, I know I'm standing like Habib said 10 minutes before the lunch time, but you give me these 10 minutes and

I'll let you know why it is important to build a great product, but it's also also important to sell that product. What Habib built here as part of the tech transformation in Lowe's is the self-checkout experience. While we were doing that, Lowe's was going through a massive tech transformation of its own. We built our own order management system, our retail media network, and our own cloud infra.

And all these products which Lowe's was working on was indigenous and was homegrown. At this juncture, we decided to create another revenue stream for Lowe's and that is how Meridian came into picture. Now, Lowe's Meridian is the technology arm which is commercializes the software which is built homegrown in Lowe's. And as part of that, the first software which we launch into market is our point of sale

solution. Why point of sale solution? Well, uh any retailer when it starts their tech transformation, it categorizes its product and the entire ecosystem in three to four different phases. The selling channels, the fulfillment systems, the supply chain, finance, payments, and auto accounting. But the most impactful and the most effort related is the selling channel. We wanted to contribute back to the industry by monetizing point of sale

solutions and at the same time, we wanted to give the control back to the retailers. And that is when we came up with with our tagline, built by a retailer for retailers. I want you to see this and this is the checkout platform which Habib has created. It is modularized, like he said, layered architecture, everything great, right? he talked about the device manager and he missed to

say some certain things here which I'll cover. Device manager was one of the most important key factor which we have created and it gives ability to be able to integrate with any peripheral device which is available in the market. Generally, the traditional POS machines comes with their own peripheral devices. And when we built that, uh we went ahead and it automatically gave me idea to not only

sell this to a big box retailer which is proficient in home decor which is Lowe's, which can also go for grocery, which can also grow for apparels, uh for quick commerce and maybe for convenience stores because if for the weighing a scale, RFID can just be added. We went ahead and added prompt engineering in that. Any developer or anyone who can answer 10 questions about the peripheral

device can actually get that device added into the system and can roll out across the chain in an hour time or so. This gives the great flexibility to the retailer to go to the any pet device, for example, what Habib was talking about if the device goes for a toss. But additionally, when we started this journey of Meridious, there were three questions which we were thinking. What

do we have? What do we need to build? And how do we sell it? We already had a great product, what we need to build was to externalize this system which can be suitable for any of the retailer market. We had a robust foundation. We had a great back end, but we wanted to build something which a customer can touch and feel. And that is when we

pivoted and use our control center. This control center, which we have built, it gives a retailer the ability to have the full visibility and monitoring to each register and peripheral device across the chain. But it also gives you a persona-based config management and ability for convertible lanes to manage workforce for small retailers. A big box retailer may not be too worried about the workforce, but a small

retailer wanted that and it wanted it in the fingertips. And that is how control center came into picture. when we were talking and selling it to different customers, the biggest thing which we realized is the maximum effort goes to a customer is on the integration of that POS machine with the retail ecosystem. And that is when we digged into the software suite which Lowe's was producing, and

we found something called as commerce gateway. This product came into existence 2 years capturing orders from social media sites and capturing the order from the customers right when the inspiration starts. We attached our APIs to those social media site, and this commerce gateway created a wrapper which insulates the internal systems in the Lowe's to the outside market. We bundled this commerce gateway with our POS solution and

gave ability to the retailers to connect with this machine in 3 weeks to 4 weeks time, which would have taken years or maybe quarters at least. the fourth building block um which Habib was showing around the checkout app, so we used the same robust foundation and we ventured into mobile POS, mobile checkout, omnicart, and shared cart. So, if you look at it till this time, we are

covered with our robust foundation and flexible architecture. But what we have still not got is the X factor. And for that, we went to US and we found a lot of industry standard items where few folks were using Vision AI to be able to identify an item in a tray or they were also finding customers scanning the items while they're walking down the lane. So we used

and we created our own Vision AI. We implemented convolutional neural networks and trained our AI models to start identifying loose items. Which means that we were doing an accurate identification of a Washington apple in with respect to a Royal Gala apple. And our accuracy was such a big huge hit when we went into another and we'll talk about it. Uh it was one of the most accurate

product which we have done. And that is when we started to leverage Vision AI for multiple other things. All the products which you are seeing over here which is Vision AI based item identification, label swap. Label swap is where, you know, a high value item label is swapped with a lower value item and there's a shrinkage which can happen at the center. Real-time data analytics which means

a customer is walking across the store. You can see if the customer is stopped at certain location, you can go and help that customer and I actually upsell the product which he's looking for. And then similar to that, we also created our store companion app which was developed which was uh developed over the MCP servers. And this was giving the real-time transaction details to the store managers,

the business heads. If you find any unusual sales pattern emerging from a particular register we were using this store companion app to do this. we were using the same technology across the board to be able to solve multiple bigger problems. And now I'll culminate our points with a single agenda which I think is the most important one. When we went to NRF, we found that the technology

by far was common across the board, right? You guys have gone to Decathlon, you have experienced the RFID technology over there. Um you have seen how easy it is to check out from a Decathlon because of that RFID technology. But can we use the same RFID technology for expired items in the shelf? Can we use the same RFID technology for damaged items during transport? Can we use

the same RFID technology for maybe the supply chain optimization or backroom operations, What we saw in NRF 2026 is the technology by far remains the same. Generative AI models are the conversational starter these days. These are not the moonshots anymore, of course. But the idea is what problem you are solving. And as a developer, what we need to understand is that everyone over here using AI can

actually develop and can start coding. But you need to have the vision of this product to become the enterprise. It cannot be built for a project. Always build a product which is actually from enterprise mindset. And that is where I want you guys to leave with four key takeaways. Number one, whenever you start building a platform with an enterprise mindset, don't solve it for only one project.

That is what Habib did when he was creating the self-checkout machine. That is why we were able to use that self-checkout machine, make it externalize, and go into the market. Second, try to build a product which can be sold in the market. It automatically becomes the best product. If you are building a product only for that project or within that organization, you may have a good product,

but you will have the best product on its own when you will think about the industry when you will build it. I talked about the RFID technology. Now, this is what exactly I mean here. The tech remains the same. It's about how we solve problems that shape millions of customer experience in everyday lives. That's our responsibility. If RFID can be used for checkout, it can be used

for multiple other things. If vision AI can be used for package item identification, it can be used for label swap, it can be used for uh identifying customers in need in the store. Use the technology for the right business problem. And the last one, I think uh this is something which I'm now I'm talking to the engineers who are in the early stage of their careers and

below of 8 to 10 years, right? Uh the most important thing which I think is in this age of, you know, AI, uh the platform you build defines your capability. And how you build it defines your craft. And everyone will learn coding. Everyone will be able to do that after some time. How you sell it in the market defines your legacy. So, if you are good at

building a good product, you should also be good in selling that product. When I say selling that product, I don't mean it's always to be the customer. I mean it it can be your leader. It can actually be your peers. And it can be your team as well. And this is where I want you to leave. And thank you. Thank you, everyone. >> [music] >> Woo!