Businesses and organizations are filled with all kinds of data that could provide valuable insights into their operations and decision-making. But first they need a strategy and the proper tools and technologies in place to put it all in motion.
In this episode, we look at various data-driven strategies, challenges, and best practices from across industries. We also explore the key components of a successful data-driven culture, such as collecting and managing data, leveraging artificial intelligence and predictive analytics, and forming partnerships with other businesses. By understanding these concepts, businesses can harness the power of data to unlock insights and make informed decisions.
Our Guests: Awiros and Vistry
Before founding Vistry in 2020, Atif worked as a strategic advisor for Accel Robotics, and was Vice President of Teradata. At Vistry, he focuses on providing restaurants with an intelligent platform to measure and improve their performance.
Saransh has been with Awiros for almost five years—in various roles such as Lead AI Engineer and Computer Vision Engineer—where he’s worked to build an ecosystem around the company’s video intelligence solutions.
Saransh and Atif answer our questions about:
- (1:59) The meaning of data-driven cultures
- (6:47) Technological advancements making data more accessible
- (8:45) Successfully implementing a data-driven strategy
- (12:20) AI’s role in creating data-driven cultures
- (17:27) Lessoned learned from the industry
- (24:41) Leveraging expertise from partners
To learn more about creating data-driven cultures, read The Business Value of Data-Driven Cultures.
For the latest innovations from Awiros and Vistry, follow them on:
Christina Cardoza: Hello and welcome to the IoT Chat, where we explore the latest developments in the Internet of Things. I’m your host, Christina Cardoza, Editorial Director of insight.tech, and today we’re talking about data-driven cultures with a panel of expert guests from Awiros and Vistry. But before we dive into the conversation, let’s get to know our guests. Atif Kureishy from Vistry, I’m going to start with you. What can you tell us about yourself and your company?
Atif Kureishy: So, our company is two and a half years old. We’re a startup focused in the restaurant-automation space. I’m based here in Southern California, and we’ve really enjoyed our partnership with Intel and working at the edge, IoT, and computer vision, and excited to be here with you today, Christina.
Christina Cardoza: Yeah, excited to have you and to learn a little bit more about what you’re doing in this space. But first, Saransh Karira from Awiros, please tell us more about yourself and the company.
Saransh Karira: Hi, Christina. So, basically I head engineering at Awiros, and we are a startup based in India. And what we’re working on is building an ecosystem on top of a platform for radio-intelligence use cases.
Christina Cardoza: Great. So, I don’t think it’s a surprise to anybody listening or to you both and to everyone in the IoT landscape these last couple of years that we’ve all been talking about how data is so important to business success today. It has all of these valuable insights that we need to gain and reach. But one thing that is sort of new is that businesses and organizations are now creating their cultures around data and talking—like we mentioned—we’ll be talking about this idea of a data-driven culture.
So I want to start the conversation there, and, Atif, I’ll start with you. When we say data-driven cultures, what are we really talking about? What’s the importance of this, and why is it something that businesses care about so much?
Atif Kureishy: Yeah, for me I think it’s simple, which is when I hear data-driven cultures it’s really about making decisions that are evidence based. So ones that are grounded in the understanding of data coming from your enterprise, being able to trust that data of course, and analyze it and derive key understanding from it. Then ultimately making decisions that drive strategic advancements, strategic initiatives, what have you. So that is kind of a broad-base statement.
You know, if I think about where customers have gone over my career I’ve worked with many Fortune 100 companies, let’s say, all over the world. And when I look back, first generation of a data-driven culture is really about data acquisition and data understanding. Can I acquire all the data that I potentially have access to inside my enterprise or outside, and then report on it frankly and just make sense of it to consumers and leaders that are interested in that information?
I think the second phase of that journey is kind of what we’ve been on for the last decade or so, or two decades, which is then starting to do prediction on top of that. And that introduces a lot of concepts, let’s say, in the machine learning space, if you will, around how to do that effectively and be able to explain and build confidence in those systems. And then really I think we’re on the third generation, and what the next gen is starting to look like is really with the introductions of these large language models. And rather than having very human data science, data-engineering-intensive activities, now moving towards really AI-based systems that tend to be smarter than us. And so how do we share a large corpus of enterprise data with those LLMs and do that in a trustworthy way to still make decisions that are informed in the enterprise.
So, somewhat of a long response, but that’s more from the technical perspectives. There’s for sure a people and organizational element of establishing that culture, but that’s, at least from my perspective, what I think of.
Christina Cardoza: I absolutely agree—those different evolutions or generations we’re seeing, it all started with computer vision and being able to see more into your business operations. But then we keep evolving real-time, artificial intelligence—just how we exactly see into those operations and how we’re collecting and managing and analyzing all that data, it just continues to change. And one thing that I think is really interesting about the data-driven-culture aspect is that it’s not just one industry we’re talking about here.
Vistry, I know you guys do a lot of work in the restaurant- and retail-automation space, but it’s really every business can benefit from data-driven cultures. And Saransh, I know you guys have a focus on video AI and enabling all of these AI applications across the various different businesses and industry. So I’m curious about what you’re seeing on your end and the customers that you’re working with, how they’re viewing this idea of data-vision cultures.
Saransh Karira: Yeah. So, I think in the last three to four years we have seen tremendous changes in the landscape. So let’s take an example: earlier there was the—data policies were like an umbrella term for any kind of data. Anyone if they hear that the data needs to be accessed for, let’s say AI training or anything, they were like, No, not possible.
But now all the customers—our customers, the business leaders—all the people are themselves becoming aware, and they know that the amount of data that you give to the system is the amount of precision that you get from the system. And so the data policies are now—now have clear boundaries, and the other side of the boundaries are very valuable for us, the amount of data that we can get from them.
Christina Cardoza: Absolutely. And I mentioned, Saransh, your company works with video AI, so obviously there is a video component that goes into it. We talked about the various different evolutions and generations of computer vision, so I’m curious to hear what advancements or evolutions are really making this possible—making data more accessible, making data more valuable, and just making businesses more easily able to succeed with these computer-vision AI applications. So Saransh, I’ll start with you again on that one.
Saransh Karira: Yeah, so I think first is, again, the changes in data policies: they themselves make the data a lot more accessible than they were before. So the raw data is the first step towards it. And now once we have this raw data, now applying intelligence on it just makes it more palatable. So now let’s say you have thousands of hours of data, like you just get an information overall even when you have access to the data, it’s not really accessible, like you cannot sift through it.
So that’s where the systems come in, the intelligence systems, the machine learning systems—all of those things come in and now, like Atif said, with the introduction of LLMs that’s just more palatable. I know one of my friends, so he’s working on a product; so what that does is you just train the product on the whole corpus of your laptop and you can search whatever the documents are. You just ask the questions and it will give you the answers. So, yeah, it’s changing very rapidly.
Christina Cardoza: Yeah. And that rapid change I assume can sort of create some challenges or complexity for businesses trying to implement these data-driven cultures and these strategies and efforts. You know, there’s so much data coming in now they have to be able to get the right information at the right time; they have to get real-time information; they have to maybe decide when and when not to store data to see the historical patterns or make predictions in the future. And so there’s a lot that goes into it: the technology, the camera systems, the AI, the edge, the cloud. So, Atif, I’m wondering what are the challenges that you really see businesses face. Where are the hurdles, and how can they overcome them?
Atif Kureishy: Yeah. Just a reminder—we are focused in the restaurant-hospitality space. So naturally when you think of that customer, it is a very people-oriented business, high velocity, and relatively unsophisticated. If you think about a restaurant environment, they are starting to make a lot more technology investments. But historically that’s not been the case. So, what we’ve seen in the challenge, number one, is that any type of capability that gets deployed and scaled across a large number of locations has to be very cost effective.
So this is where I think especially Intel has brought a unique value proposition, in the sense of you can run on commodity compute that’s been in existence there in the restaurant, or potentially deploy next-generation compute and have machine and deep learning models that can run effectively there at the edge. We position the Kubernetes runtime there, and so we have a lot of flexibility to provision and deploy different, let’s say, inference workloads that are running on camera-based data or other types of sensors.
And a lot of the things that we’re tracking are objects in the kitchen. So it makes for a unique environment, and for sure our training infrastructure has to be robust to be able to detect and track and be able to understand the activities that are occurring in the kitchen. So I think some of the key challenges in all of that—I mentioned the cost element of ensuring that that edge is robust and can be managed, let’s say, from a cloud-based infrastructure, and that you can get consistency across thousands of locations. And, again, that’s where some of the technologies around OpenVINO and deep learning tools that the Intel group has provided have helped tremendously. So, we can run our inference workloads on, let’s say, Intel Atom® tablets. We can run on, let’s say, i7 Tiger Lakes. We can run on the new Alder Lakes very easily and be able to optimize those runtimes effectively. So that’s been incredibly useful for us and for our customers.
Christina Cardoza: Absolutely. And I know you’re in the restaurant space, but I think a lot of what goes on in the restaurant or what you’re looking at in the kitchen really can relate to other industries—worker safety, inventory management, defect detection—these are all things that in the manufacturing space or smart cities, all businesses want to get insight into and to start seeing where they can make changes; what’s happening.
And a big part of that is artificial intelligence. These new AI models are making it happen, are being able to detect these in real time, and alert managers or operators at the right time and not give like a whole bunch of false notifications. They really are able to see, “This is something that I need to address immediately,” or, “This is something that I can see over time that I should discuss with my workers.”
So, Saransh, I’m wondering, because you work with a lot of AI applications, how are you seeing businesses being able to approach AI? And what really—as part of this puzzle of this data-driven culture—what really is the value of AI in these efforts?
Saransh Karira: Yeah. So, what we are seeing currently—first of all, again, the data policies are changing, and because of that a lot of infrastructure is being built for integrating a lot of data. So what I think the value and the value of data is is when you can connect a lot of different types of data. So, if you can take each data as a dot, and then if they can connect together the sum is more than the parts. So a lot of our customers are, let’s say, connecting throughout their different infrastructure or their different divisions, and across that you can go to one place and you can just get access if you are from—you belong to some other division, for example.
So that is one of the use cases, but it extends to a lot of different organizations—even we work with government extensively, and what we are seeing currently is they are connecting the vehicle preregistration with the cameras and then the passports. Everything is getting connected, and then the data becomes, the interconnected data becomes, much more valuable than the one system that is standing in just a silo.
Christina Cardoza: And I think a part of this data policy that we keep coming back to is data privacy and making sure, especially when you have cameras on people and you’re collecting all of this data, that we’re doing it in a safe way, that we’re not using the data in any other way than what the businesses intend to use it for. And I can assume that this is particularly, not challenging, but an issue that concerns, that people may have—especially in the restaurant space when you’re tracking people in the kitchen or you’re tracking customers. So, Atif, I’m wondering, what is—your history is AI data strategy—how do you ensure the privacy and the concerns of it?
Atif Kureishy: Yeah. So, first of all, I do, as Saransh mentioned, government has been on the forefront of at least integrating large amounts of data and ensuring the privacy and security, some better than others. I do have an intelligence background; so, I did work for the US intelligence community for several years. And I would say that that was a very large focus of ensuring that data handling and data privacy, data sovereignty, all of that was managed effectively.
To the question of in the restaurants in particular, we as a company don’t use any biometrics. So from that perspective there are no biometric features that are sensitive, and we’re very careful about any PII data that is collected. But generally we’re looking at objects—these may be vehicles, they may be people, but they’re not—they’re anonymized in some ways, and tracking of food products, of equipment, utensils, those types of things. So the data concerns are less, but they are still there for sure. And that workload and data storage and data transit in terms of what resides at the edge and what comes back into the cloud is thought through very carefully.
So, one aspect that I’ve mentioned several times is the inference—how do you actually get the answer of what you’re seeing and be able to react to that? And as you said, Christina, a lot of that is in the eventing or alerting or whatever you’re actually focused on. But then there is the aspect of training and fine tuning and taking that data and making your models more intelligent. And so that needs to be thought through carefully, because usually that’s acquired and consolidated, let’s say, in a cloud-based infrastructure to then support a model-retraining effort, and the data-security elements of that also need to be considered, very much so.
So in the most generalized form, if you’re not processing biometrics and you’re not looking at and storing PII, the problem domain becomes a little bit more straightforward. But data privacy, data access, just like in any SaaS-based product, has to be thought through very carefully.
Christina Cardoza: So, I’d love to hear more about how this all works in a restaurant space, or how you’re using AI and all of these technologies that we’re talking about to really create data-driven cultures and strategies for the customers that you help. So I’m wondering if you have any use cases or customer examples you can share with us: what the problem the customer was having, how you came in and helped them, and what can others learn from the challenges that you worked through?
Atif Kureishy: Yeah, a lot of what we see is these are dark spots, if you will, or parts of the business that are opaque to an above-the-restaurant leader. So let’s take the example of production control, which means you get—a restaurant essentially is a mini manufacturing site, I think, Christina, you said that earlier. And you know previously before starting Vistry we had worked in the industrial-manufacturing space, and so a lot of these concepts apply very directly. In a manufacturing sense you have measurement of inventory and you have QA and oversight of work products. And so if you apply that into a restaurant space, imagine that you have orders coming in; those orders can come in through digital, those orders can come in through the drive-through, they can come in through dine-in. And when those orders get acquired, let’s say they get consolidated into a kitchen that needs to essentially turn those orders around and manufacture, if you will, the orders correctly.
Now some of the areas of where AI is coming into play is can you create a production schedule like this is in the sense of a quick serve or fast food restaurant, where they pre-make certain products and they hold them, okay, so that when you make an order—as long as that menu item is being held in compliance, meaning it’s still fresh and it still doesn’t have any food safety issues—that is the ideal scenario, because you get your food as quickly as possible and that menu item ideally has been made in a compliant way. Where AI and ML come into play is how do I build and manufacture those menu items efficiently by predicting how many inbound orders I’m going to get and what type, and that allows the kitchen to be much more efficient, and not only from a labor perspective but also from a food-waste perspective.
The other aspect of where we’ve been using computer vision is on inventory management. And so having cameras that can look at a bowl or a pan and do volumetric estimation of how much product is in those pans can then help to inform a production schedule to say, “Hey, when there’s this many servings remaining and I predict this many new orders coming in, tell the cooks to start cooking more.” And that, again, from a lean-manufacturing perspective, that’s sort of like the just-in-time concept.
So those are some examples where you can start to look at computer vision and, again, that’s more from the supply side—so, modeling demand and then using AI to essentially ensure that the supply is there. Other mechanisms of computer vision is starting to use cameras to see how long the queue lengths are, both in the drive-through and the dine-in. And that is really about doing the demand modeling. So if I can see that there’s a stack of 12 cars, I can expect obviously they’re all going to place orders, I can then take that and input that into a prediction model and start to anticipate potentially what they’re going to order. And that is, in essence, how the optimization of the restaurant is taking place to be more data driven.
Christina Cardoza: Yeah, I love that example because when we talk about these data-driven cultures, it’s not just one aspect of a business’s operation, it’s all of these different aspects that you mentioned connected together to really create the best business flow and value as possible. You mentioned supply and demand, making sure you have all the inventory there, making sure that you have enough food to get through the day or to serve all the customer’s needs. And then making sure the kitchen is cooking the food, all the way out to customer service and quality, making sure food is fresh, it’s fast, they’re not waiting for it, and it’s the correct order. So it’s really an end-to-end production line connecting all these together and using data to drive all of that.
Saransh, I’m curious what kind of—outside of the restaurant space—if you could provide any customer examples or use cases of what you’re seeing, how you’re helping solve the pain points that you have dealt with with your users, and how others in the industry can get over some of those pain points.
Saransh Karira: So, over the years I think we have seen a lot of these use cases and a lot of these surprise ourselves as well. So, they extend across the different industries, and one use case was, I think there was this—basically there was this deployment where there were multiple different campuses, and for each campus there were multiple different access points or access sites. And the original implementation was just to see how many people are coming in, how many of them are visitors, how many of them basically have the access to the site, or how many of them are the first-time comers, so on and so forth. And that was the initial use case.
But the customer managed to use that to basically change the configuration of their security personal depending on where the people were; where the crowd fall was more they basically changed the security there and they reduced it from the other access points. So that was very interesting to see.
And other than that, I think we have seen a lot of these, basically what we can call meta-analytics; we have seen a lot of these in retail. So, in retail we have seen that basically the customers use that for, let’s say, there is a point where a lot of people are coming in. So it basically generates a heat map where the footfall is more, where it’s less, and depending on that our customer basically can change the configuration and the placement of the things, inventory management, and so on and so forth.
Christina Cardoza: Great. And one thing that comes to mind when we’re talking about all this, all the technology and the intelligence that goes into this, Atif, and you mentioned Intel a couple times throughout our conversation. I should mention that the IoT Chat and insight.tech as a whole, we are sponsored by Intel. But I am curious, because it always seems like when we are trying to create data-driven cultures or make things happen in the industry, these days not one company can do it alone. It’s really working with others and creating an ecosystem and leveraging technology and expertise from different companies to make this all successful and possible. So, curious how you work with partners like Intel. What’s the value of that, and how does that help drive, in this situation, data-driven culture? So, Atif, I’ll start with you on that one.
Atif Kureishy: Yeah, happy to answer that. You know, we are very thankful of our partnership with Intel and, as you mentioned, it takes a village, or it takes a broad ecosystem. So, around ODMs and OEMs that are providing the Intel base compute, working with the system integration teams out there that ultimately need to place these edge devices and sensors at the locations so that this processing can occur.
And then of course, having a cloud-based infrastructure, we work very closely with AWS, and so Intel is a key part of facilitating those dialogues and interactions with that larger community. And then of course the robust set of tooling and infrastructure that’s provided really around OpenVINO. So that’s been, that’s all been great for us. And it allows us to optimize, again, the types of processing that we’re running on CPU or on the IGPU—integrated GPU. There’s also good support of working with the open source community and the various deep learning frameworks that are out there. So that has been wonderful.
Christina, I want, if you would allow me, I would like to just go back to the previous question that you mentioned around data-driven cultures and use cases that are in the restaurant example, and we kind of threw around a few of them. I wanted to highlight what the historical culture is in the restaurant, because I think it’s important to understand that and how it makes sense that we’re now using data to serve the customer more effectively. And because the ordering infrastructure and loyalty—all of that is being generated digitally—it’s becoming easier.
But if we think about 20 years ago what the culture of a restaurant was, it was really reliant on the people and the managers that were running the restaurant and using their intuition of, “I expect a lunchtime rush today,” being aware of events that are occurring, catering events that are occurring—we’re expecting a field trip to come to the restaurant and 30 kids to show up, and we have the breakfast usuals that come in, and here’s how I’m going to place people.
So that’s been the last couple of decades of what working in the restaurant is, and especially with the pandemic all of that got turned upside down, because dine-in wasn’t the priority. Of course drive-through was still a key part of it, but digital became more essential, and now the kitchens and these restaurants are inundated and overwhelmed with trying to fulfill these orders. And so there is a need to essentially serve their customers and do all of this in a data-driven way.
And that is—I think that’s been the phenomenon that we’ve seen, which has been really exciting. So I just wanted to highlight that for the users, is that that is traditionally how—and, by the way, there’s a ton of restaurants, especially small restaurants and local restaurants, that still run that way. But when you look at the larger brands, they’re moving absolutely to more of this data-driven culture.
Christina Cardoza: Yeah, absolutely. And that’s a great point. Especially when you’re talking about quality of food and customer loyalty with things becoming digital, sometimes the customers never make it into the restaurant, or sometimes a third party is delivering their orders. So it’s that much more important that restaurants are on the top of their game, that they’re able to provide all of these services and make the customers happy.
So I think this data-driven culture is, as we move forward into the future, is going to become more and more valuable now that customer expectations are changing and the demands on businesses like restaurants are changing. There are going to be a lot of evolutions coming rapidly, I think, to some of these industries that haven’t taken advantage of some of the technology—now they’re really seeing the benefits of doing so.
And I think all the Intel technology that you said you were using, that’s really helping move all of this along—OpenVINO with the AI. And I think working with a technology giant like Intel, it’s also important not just for the technology that they can provide, but the partners and the ecosystems that we talked about that they can open you up to and connect you to others. So Saransh, I’m wondering how you’ve been working with Intel or other partners in this industry to make some of your solutions or use cases happen?
Saransh Karira: Yeah. So, as we’ve mentioned about ecosystem a lot of times in this chat before. So, I was also talking about with our platform we are trying to create an ecosystem of video-intelligence applications. And for that, the lower—so basically it starts with the hardware and it goes to use cases and then it goes to the marketplace. So the hardware is where Intel comes in. And then on top of it there are different use cases that are being developed by different researchers or any of the third-party developers, anyone. And on top of this there is a layer of marketplace which can be—which is basically visible to the end customers.
So in this hardware layer, the Intel community, the hardware, the software—all of these things have helped us tremendously. So, I think at the edge Intel is very cost effective for us, first of all. And the libraries have helped us a lot in optimizations. So there are different various amounts of optimizations, be it for inferencing—so, the actual part where the AI runs—as well as the decoding part of the video, and many other things. And Intel provides specific hardware for different operations like video recording and inference and all of these things. So, and also the support is very, very, very wide. So that’s what I think is where Intel has helped us a lot.
Christina Cardoza: Great. Well this has been a great conversation. Unfortunately we are nearing the end of our time, but before we go, we talked about different evolutions that have been happening in this space. I think that those evolutions are only going to continue as time goes on, as business needs change. So before we go, I just wanted to throw it back to each of you, any final thoughts or key takeaways you want to leave our listeners with when it comes to creating data-driven cultures and what the future of this looks like for businesses? Atif, I’ll start with you.
Atif Kureishy: Yeah, we touched on it a little bit. Really the GenAI space—ourselves, like everyone else, has sort of gotten on that bandwagon, if you will, and really worked extensively with models like GPT-4 for the last several months. And what’s interesting for us is a lot of our focus for the first couple of years has been generating, let’s say, dark data. How do we apply computer-vision workloads at the edge to create a data stream of physical observations? That’s really what we’ve been doing.
And that data then needs to be stitched into a larger baseline or foundation of data that’s coming from the point of sale, coming from inventory-management systems, coming from time-reporting systems, and so on and so forth. And so we’ve been looking at LLMs, large language models, to really interact with a larger and broader set of data and make sense of it. And the ability to do that very quickly is really fascinating and phenomenal. So, like nothing I’ve seen—especially being in the machine learning space for the last decade or decade and a half, it’s really exciting what the future looks like and how to very quickly position these capabilities to solve distinct problems.
So if I were to leave this audience with something is to—beyond ChatGPT and getting recipes and looking for travel itineraries and generating poems, which I’ve done with my kids and we have a lot of fun doing that—but it does really have big implications into the enterprise, and we’re excited to be a part of that journey.
Christina Cardoza: Absolutely. And I can’t wait to see what else Vistry comes out with over the next couple of years. Saransh, any final thoughts or key takeaways you want to leave our listeners with today?
Saransh Karira: Yeah. So I think my thoughts are very much in sync with Atif. So I think this new view of AI that has been coming and from the past few months is really exciting, and we should really watch it out what’s happening. So, there is a lot of different things that’s happening in the performance sector and LLMs especially, and a lot of generative, different AIs, and when it all works out and connects I just want to see what happens. So, yeah.
Christina Cardoza: Yeah, absolutely. And I would urge any of our listeners, if you’re looking to create data-driven cultures, get your efforts off the ground; you don’t have to do this alone. I would urge you guys to visit Vistry and Awiros websites to see how they can help you out because, like we mentioned, there’s a whole ecosystem out there and a lot of knowledge and help beyond that. So, just want to thank you both again for joining the conversation. It’s been very insightful and informative. And thank you to our listeners for tuning in today. Until next time, this has been the IoT Chat.
The preceding transcript is provided to ensure accessibility and is intended to accurately capture an informal conversation. The transcript may contain improper uses of trademarked terms and as such should not be used for any other purposes. For more information, please see the Intel® trademark information.
This transcript was edited by Erin Noble, copy editor.