[00:00:00] Luke: You’re listening to a new episode of the Brave Technologist. This one features Ashita Ghosh, who’s the senior manager of data science at Walmart USA. Ms. Ghosh is a skilled leader in data science and machine learning, combining technical expertise with strategic insight. With a master’s in data science, she excels both in research and practical application.

[00:00:15] At Walmart, Schulenberger, and AbbVie, she drove innovation and delivered results in computer vision, machine learning, deep learning, and generative AI across manufacturing, safety, and operations. Her active research contributions and leadership at major organizations highlight her significant impact on the field.

[00:00:31] In this episode, we discussed AI innovations at Walmart, including generative AI for employee assistance and inventory management, challenges related to regulatory compliance and the role of human oversight in AI applications, and how we can use AI as a tool for creating a better world rather than thinking it’s there to take human jobs.

[00:00:46] And now for this week’s episode of the Brave Technologist. Ishita, welcome to the Brave Technologist. How are you doing today?

[00:00:57] Ishita: I’m doing great. And thank you so much [00:01:00] for inviting me here. I’m so glad to be part of this and I’m so excited to chat with you.

[00:01:05] Luke: Yeah. I’ve been looking forward to this one. We’re here today at the AI Summit in New York.

[00:01:09] And I know that you’re, giving, I think at least one talk here, right? At the conference. Would you mind sharing with the audience a little bit about what you’re talking about at the conference today?

[00:01:16] Ishita: Yes, absolutely. So I’m really glad to share that today. I have two sessions when I completed this morning.

[00:01:22] It’s about a I chronicles and the obstacles off a I for tomorrow. So that was a chat. Really interesting with my moderated Christie Mello. And then it was room full of audience and everybody was very engaged. The next session will have about data dreams and AI dreams, which is like, uh, I think five to six member of panels.

[00:01:44] And then there will be a moderator as well, who is very well known kind of AI ethics, kind of person. So I’m really looking forward.

[00:01:52] Luke: Excellent. Excellent. That’s great. Walmart’s emerged as a leader in AI readiness, implementing kind of generative AI for [00:02:00] employee assistance and inventory management.

[00:02:02] Can you elaborate on how AI innovations like these are shaping retail and as an industry and kind of what lessons other industries can take from the lead that you all are taking?

[00:02:12] Ishita: Absolutely. That’s great. This look up perfectly like a good question for the starter. Before we delve into the details about that question, I’d like to like briefly introduce myself so that the audience know better.

[00:02:25] My name is Ishita Ghosh and then I have been in this AI and machine learning industry for about close to 10 years. And then I have worked across many industries, not just limited to retail. So for example, I worked, I led the AI innovation for, uh, uh, AbbVie, which is a pharmaceutical industry. And then Schlumberger, which is like the world’s number one oil and gas.

[00:02:49] Um, and then I switched over to Walmart. and before I had many years of consulting experience back in my home country. So based on all of my experience and personal [00:03:00] knowledge, I would be whatever I would be talking here today will be based on my personal opinion. And I just want to make sure that none of this is a representation of my organization.

[00:03:12] So after making sure of that. Let’s get into that amazing question,

[00:03:16] Luke: which

[00:03:18] Ishita: is about Walmart. So as we all know that Walmart is like a household brand, and then almost everybody is aware of Walmart and it’s probably the world’s, uh, it’s a fortune one company. It’s probably the world’s largest retail fleet with more than 10, 000 number of stores and then dealing with 230 million kind of customers every week worldwide.

[00:03:43] Any organization kind of operating at this large scale would definitely need to keep up. To what the customer needs and evolve into the AI space. So if I started going back to the question, I think you were more interested to know, [00:04:00] like, how is Walmart keeping up to this? Like, what are the, uh, AI innovation they’re making, right?

[00:04:05] And

[00:04:05] Luke: yeah, I mean, kind of a, Walmart’s such a big, you know, both in product diversity and, and kind of covering almost everything. Right. Like, so when you think about it from. The ability to have like you supply chains to then the customer experience and then having such a big footprint too.

[00:04:22] Like I imagine it’s just kind of like, one of the bigger challenges is just figuring out where to start, right? Like, or where to apply it or, or how you have applied it. And maybe, letting folks know kind of. what’s worked really well and what others can take from your lead.

[00:04:34] Ishita: Absolutely. So for Walmart, as the founder, Sam Walton said, like customer is the God, like customer is everything for us. And then there is one favorite quote that I have from Walmart, of course, from Sam Walton is like, he mentioned, listen to your customer. If you don’t, someone else will. So that has been like the guiding principle for all of us within Walmart.

[00:04:57] And keeping that in mind, Walmart [00:05:00] has made a lot of innovations in the space of AI, keeping the absolute customer need in the mind. So for example, previously, when we would go to the on site shopping or online shopping or anything, we would just have a search bar, right? And Walmart is trying to re innovate that with AI to a more personalized way.

[00:05:24] So I think there is a mention that Wal Mart is kind of building towards Gen AI tool. It’s called Wallaby. And it will be the first Gen AI model trained on the vast set of retail data, which is not been there before. So it will have like knowledge of everything within retail, but it will also adapt and learn to the specific customer profile it’s dealing with.

[00:05:52] So for example, imagine you have logged into your profile and all the experience that you will have there, [00:06:00] including the interaction with your Gen AI super helpful assistant would be tailored for you. So it will not be like very, very general, like, Hey, how can I help you and connect me to an agent? It will be able to do a lot more.

[00:06:15] And with that, there are also certain AI innovations that Walmart is doing, keeping in mind the competition with also the recent innovations in other companies like Amazon has made. One of them, Walmart has, uh, kind of given a sneak peek in one of the medias is, called Retina. And that is our augmented reality and virtual reality based tool, which given the capability to a customer, specifically targeting with the young generation audience or the customers that when they want to experience something without going to the store.

[00:06:53] At the comfort of their home, they can do it exactly as they can at the store so they can imagine the [00:07:00] 3D view of the furniture or any other game station. They’re looking into. They can absolutely place it at a different corner of their home and then they can experience the same. So that’s been really engaging.

[00:07:12] And I’m also really excited when that comes into the market for the customers, and we all can leverage it, right? So, from this learning, I feel like Walmart is maintaining that cutting edge in the market, and they are also being very competitive with the other retail giants that exist. Innovating in the AI space, machine learning space, and with all the keeping up with all the advanced IT infrastructure is really a way that other companies or other organizations can really look up to the Walmart to maintain their business model.

[00:07:51] Luke: How has it been for you? Like going from kind of, you were saying you were going to do in a pharmaceutical. Is that, is that right? Going from like something that’s kind of specialized, but still pretty [00:08:00] like, uh, there’s a lot to pharmaceutical space to something like Walmart, where you have like produce clothing, all these different, even pharmacies, right?

[00:08:09] Like, built in it. how is it from your role? Like kind of going from something pretty focused to something where you have all these different buckets?

[00:08:18] Ishita: Yes, it’s definitely been a journey. So I started with pharmaceutical. My sole purpose of that role in that company was to help reduce the timeline for drug discovery.

[00:08:32] So that was a passion for me at that time, which is still is because I still keep on contributing to the research space. That solution was more around the new introduction of computer vision. based solutions into that space, and I really enjoyed my time there. I was almost there about a year, and I was mostly very much in touch with my manager in that company who was equally passionate about making the solutions [00:09:00] and after that, actually, I kind of moved to oil and gas.

[00:09:04] So without talking about that, I cannot really explain the journey. Sure, sure,

[00:09:08] Luke: sure.

[00:09:08] Ishita: Oil and gas. When I moved into that, I realized that this is a industry that faces a lot of regulations around evolving in AI. So they would still use a lot of traditional solutions, sensor based solutions. A lot of non automated solutions and which slows us down into implementing some monitoring pipeline or safety of humans for the production of oil and gas.

[00:09:37] So I started that evolution within that company, which I consider like, which almost anybody knows is a giant in that sector and gradually realized you can be a data scientist. Or like a I engineer, but in the end, you need to convince your stakeholders to implement and utilize that [00:10:00] solution for production, right?

[00:10:02] So there is a transition that really happened in me that I not only limited my focus on the building the solution, but also Kind of making that solution available for the others who can be aware, who can be, you know, make trusted with the solution and try it out.

[00:10:23] Luke: Practical application. Exactly. And

[00:10:25] Ishita: I, I was tired that many solutions that we build as data scientists, it never sees the light of the actual production or life.

[00:10:33] So my motivation then shifted, not only building just the solution because. By then, I was kind of expert and in my regular life to build a solution, but then I shifted my focus. Hey, now I want to take this challenge and make those stakeholders relationship, make those stakeholders management from the thought AI thought leadership perspective and make it to the actual human benefit.

[00:10:55] When it comes to production, and that’s exactly when I came [00:11:00] across my role in Wal Mart when they were actually looking for the similar kind of motivated person who would not just build the solution, but also make sure that the customer, the end people. Are making to use those solutions. So even though it was like a huge jump from the industry perspective or the number of stakeholders or the number of segments that I manage in Walmart right now, but from the motivation perspective of the nature of the role perspective, it’s still aligned with what I was doing in my previous.

[00:11:34] Luke: So in this kind of journey, all these different examples are established industries, right? That have. Really good fit market was the notion of bringing a I into these work streams and in these flows like was it something you had to kind of sell. within these orgs or, or something that you proved to them and then they kind of converted over to it.

[00:11:56] Or, well, I’m just curious how it transformed things within these [00:12:00] organizations. Like were people convinced from the data sets that you were able to, or the outputs that you were able to show them from this technology when it’s applied to these, you know, traditional businesses that, Have been generating revenues for a very long time, right?

[00:12:13] or solving problems, right? Like creating drugs or whatever, right?

[00:12:16] Ishita: Yeah. So that’s really a good observation Luke, because that’s really a part of the big challenge that we face in such large traditional organizations, right? So I’ll tell you, like, there are many baby steps that needs to happen in when we march towards that.

[00:12:32] If I decide today that, hey, I know this solution is good. I know this is going to work, but I just still try to convince the, you know, other part of the organizations that saying, Hey, can you deploy the solution and make it available for people? It’s not going to happen overnight. And that’s according to my personal experience that any time as data science enthusiast or a enthusiast.

[00:12:55] As a transformational force, if we, and this is [00:13:00] mostly applicable to, and generally any traditional organization or traditional industry, right? Which faces a lot of regulations. Anytime we do something transformational, it takes a long time to make it rolling towards the production or towards like, make it available for the people.

[00:13:17] But what, Personally, I have learned from throughout my experience is building trust is not just showing how accurate is your model or how accurate is your solution. Building trust also means that to be able to show your your stakeholders that your customers are adopting towards. So you can convince them to go, you know, live.

[00:13:44] In a couple handful of stores or a couple handful of production lines, depending on your industry or a couple handful of drugs, depending again on your industry and when it is available, you try to educate them. There is a lot of change [00:14:00] management involved, and when you do that in small number of sets, you can show that it is being received well.

[00:14:07] And if the adoption number goes high, That’s really what the stakeholders management cares about, right? So when your stakeholders is seeing that, okay, even if it’s handful number of stores, it’s getting adopted. I am seeing not any more resistance, but I’m seeing enthusiasm, then they would be even open to the idea.

[00:14:28] Can we expand that? You need to get into that door. Once you open up that idea, can we expand that? Then comes your logic, then comes your data, then comes your All the reasonable insights, then you can produce them, produce those insights in front of those stakeholders, and probably that will be received well.

[00:14:48] Anything before that, it’s not gonna, it’s very hard for the stakeholders to see anything on the data other than just knowing that it’s being received well. So that’s, that’s [00:15:00] usually the trend that I have seen from my personal experience.

[00:15:02] Luke: That’s like super insightful because I mean, I think a lot of people are wondering like there’s, there’s both this kind of pressure to kind of get, get AI to be more ubiquitous or integrated, but a lot of people don’t necessarily know the best path to making that happen, especially in organizations where you’ve got people, uh, product management, things like that, that are entrenched in doing things a certain way.

[00:15:24] Or, uh, might be resistant to change or skeptical or whatever. I think your points about starting in a certain area and kind of proving it out like, or, or things people can use, uh, when they’re, when they’re trying this out, I have a question too, cause these different areas we’re talking about, like some of them are really highly regulated on their own before you even throw AI into the mix.

[00:15:44] Like how is it navigating, innovating with AI, with regulatory conditions that are already there and then. Regulatory attitudes that are both global, but also might be regionally influenced [00:16:00] like in Europe or elsewhere, like has that been a real barrier in your experience or or has it been something that.

[00:16:07] You’re keeping peace of mind, a top of mind, but are still kind of going forward. How much has that influenced your work?

[00:16:13] Ishita: That’s a really good and also complicated question. It’s a complex thing. It’s, it’s really depends on what industry we are dealing with, right? And in majority of the pharma, healthcare, retail, everything has like FDA regulations, ISO regulations for oil and gas.

[00:16:34] So it’s, it’s, it’s, It’s really important to comply with them. But it’s really interesting to me from a point of view that I like to see AI as a friend of human. But while we do that, it’s important to keep things under control and it’s important to take measures so that it doesn’t get out of control. So I believe and I feel that forming all the ethical [00:17:00] committees, governance committees and the trust committees, legal committees, to To carefully examine this regulations and try to comply with it, make a very good kind of legal team along collaboration along with the development team to make sure like we are not doing any data breach.

[00:17:19] We are not breaching any data privacy governance area and then we are taking every measure possible to, you know, cyber security angle and every angle so that the people feel they are secured with their data. That is really important. Now, it might be coming in a perspective or in a way it might come across that it’s being a hurdle or it’s being a problem.

[00:17:45] Maybe it is, maybe it is for the time being that it’s slowing us down and There are some extra steps that we need to clear before we think of a solution being deployed. But I feel in the long run, [00:18:00] when the ultimate goal is to have a solution that assists human and doesn’t breach the human trust, those steps are necessary.

[00:18:08] So I look at it more of a long game rather than just try to overcome certain stages very fast to try to get something out of the door. But it’s more beneficial, according to my opinion, that you comply with all of the stages, even though it’s very complex regulatory advice. And then you make a foolproof, trustworthy future plan before you get something into the market.

[00:18:32] Luke: Makes sense. And I think it’s one of those things where there’s so much education that has to go on, like both like the within the regulatory space, right around how the technology works, but also within the organization too, right? Like, I would imagine that. Have you seen that kind of help bring teams together in a way by educating them and showing them how this works?

[00:18:54] Like, I mean, a lawyer isn’t necessarily thinking about AI, but then once they start thinking about it, maybe it’s [00:19:00] affecting how they use their work. I mean, how they use AI in their own work, right? Like, are you seeing bonding kind of happening from that within orgs that you’re working at?

[00:19:08] Ishita: Yeah, definitely.

[00:19:09] It’s again, a very good perspective wise question. I feel I have seen a lot more trust being paid or attention being paid to the AI while working to the Cross collaboration or across the cross functional team, and I would say that because to most people, A. I. Is still a black box. So when we started like five years back, it was even really hard to explain what’s the outcome of that white out black box.

[00:19:37] And why do we even believe in that out? Because we don’t know what’s going on inside. So I think in the last five years, a lot of explainability has come. Yeah. While explaining why the decision was made and how is that beneficial for the person to be using that. And while also talking about explainability, I have observed in my personal experience that anytime you [00:20:00] communicate about AI with somebody to kind of, Convince them or try to make an alignment that it’s going to be beneficial.

[00:20:07] Always keeping the human in the loop it’s always helpful. So anytime we talk about like, hey, this is going to be completely automated, there will always be registrations. And I personally also feel human should be in the loop because human is the major, major part, one of the like key components through because of which the AI exists today.

[00:20:30] So I believe like, even though AI is making a lot of things easier is assisting us and everything, but the actual brain, the actual intelligence that it’s evolving from is the human. So it’s really important to keep. Some layer of human touch within the entire process and not making it completely automated.

[00:20:51] And that will keep the trust alive because when somebody knows that, hey, there is stage gates, there are regulatory measures, there are human in the loop who [00:21:00] is kind of monitoring and observing and has the control. If anything goes wrong, it will not have any harm over the people that really establish a lot of trust.

[00:21:09] Mental peace, the safety, the safety net, and the trust. So those are the things that I have seen coming gradually into the picture. if we utilize them while maintain the communication and education with the other kind of different teams, that really helps.

[00:21:25] Luke: Do you feel that there’s, enough user feedback loops that are happening in AI today, or, would you like to see more of that?

[00:21:32] I mean, I’m not speaking specifically to Walmart, but more generally like with AI.

[00:21:37] Ishita: I feel, yes, the more we provide the human feedback back into the A. I. System, it’s definitely gonna be beneficial, but it’s also fundamentally it’s like teaching a human, like the more we interact, the more we appreciate it.

[00:21:54] Take the positive feedback back. The more we also evolve, so will be the A. I. So it’s [00:22:00] definitely helpful. And then there is also the concept of reinforcement learning. There is also the concept of like penalizing the decisions if it has been wrong from the human feedback and penalizing the model for that.

[00:22:12] So it’s really important, but we also need to be very careful about when we use those data as feedback. It should be made aware to the customers that your feedback is going to be used for the model training or correction of the model engine. It’s definitely should be complying with those regulations.

[00:22:34] region specific, country specific, anything that applies to the state or to the country. I feel it’s, it’s really important not to keep the end users in the dark and it’s really important to maintain the data privacy. So anything that, that you are, that definitely should utilize for the feedback for AI to evolve, but not at the cost of like keeping the humans in the dark.

[00:22:58] Luke: No, I think it makes a lot [00:23:00] of sense and it’s great, great insight too. Thinking about Walmart’s just a snapshot of kind of like everything right even farther beyond that like what are personally like You had a really good example earlier We talked about working for the pharmaceutical company applying AI to creating new medicine things like that Are there any use cases that just?

[00:23:20] You’re really excited about in with AI that you think are going to like really fundamentally change the world.

[00:23:27] Ishita: That’s yes, I am. Absolutely. And that’s an amazing question is such an open ended question that I don’t know whether we can put a bet like that we can put the answer to bet. So it’s it’s huge, right?

[00:23:40] Luke: Yeah, yeah. Or just one example, you know, of an area.

[00:23:43] Ishita: One area I am really excited about is the computer vision and application of computer vision. I feel it’s really, really gonna change a lot of dimensions. Speaking from the pharmaceutical side, one of the major drawbacks of how [00:24:00] currently, or in the recent past, we have been, we have been practicing the drug discovery is a lot of changes.

[00:24:08] Human involvement without assistance of computer vision based AI. And the impact of that is that one drug discovery process or one drug kind of production pipeline process and getting the FDA approvals and everything takes longer, sometimes even a decade. I feel even the importance of health care and how.

[00:24:33] The entire nation or entire global research community is marching towards the cancer research and everything else. It’s really important and the impact will be much higher if we use a lot of computer vision based AI such 3D computer vision or kind of utilizing augmented reality and kind of experiment with those how we can expedite the process.

[00:24:57] That’s going to be, in my [00:25:00] opinion, one of the biggest achievement we can make. not just in terms of it’s accelerating the timeline for duck production, but also minimizing the cost because before the recent developments that’s been going on, it’s always used to be on the scientist who is working and manually changing a lot of things through a lot of trial and error methods to actually come up with a composition or come up with a final stage.

[00:25:22] But imagine having simulation tools. It’s assisting those scientists to run a lot of simulations in a matter of seconds or minutes. It’s going to really help. And if a drug is coming faster than ever enough, then it’s getting approved by FDA eventually is helping a lot of human being and healthcare being one of the major concerns of the entire globe.

[00:25:41] I think it’s going to be an amazing, outcome.

[00:25:44] Luke: It’s fascinating. Really, really excellent. You’ve been super gracious with your time and it’s been really interesting conversation. Is there anything we didn’t cover that you, you want, uh, our listeners to know about?

[00:25:54] Ishita: I think we covered the most of it. Every time we kind of just one [00:26:00] curious, uh, curious angle that I have been thinking about it recently a lot that there is a question and kind of fear from everybody’s mind that would AI take our job with AI take our future job.

[00:26:14] It really depends on a lot of different perspectives. And I wanted to share a little bit with the audience that it’s really not about AI taking our job. It’s, it’s more like AI is assisting us to a better world that we want to create and drive. So I think it’s not an independent, I mean, independent mind working that creating a better world without us.

[00:26:44] Doing something for it without us. So I just want to have like a really positive feeling about it. That AI is probably not at that stage yet where it’s going to take our job and kind of be [00:27:00] harmful to us. If used carefully and with the measures taken and advised by the, uh, you know, advisory bodies of the around the world, I think it’s really not in that stage yet where we should be losing trust in AI.

[00:27:14] So I would encourage the audience that they keep themselves up to date with all the recent advancements and what really makes AI responsible. From all the organizational standpoint, as well as the research community standpoint. So yeah, that’s something that’s really

[00:27:30] Luke: interesting. AI is power tools as opposed to replacements, right?

[00:27:33] There are something like that. No, that’s fantastic. If people want to follow your work or what you’ve got to say, are you out there online or is there somewhere they can go check you out?

[00:27:42] Ishita: They can definitely follow me through on the LinkedIn and then I have a personal website which is underdeveloped yet now.

[00:27:50] I didn’t get time to finish it. They can follow me there or they can follow me on Instagram or somewhere. I have a Google Scholar profile, so it’s already linked with my LinkedIn, so I guess the best place [00:28:00] is LinkedIn.

[00:28:00] Luke: Awesome. Well, we’ll be sure to link that in the show notes. Thank you so much for coming today.

[00:28:04] It was really fantastic conversation and I think our listeners have learned a lot about it and I really appreciate you giving us the time.

[00:28:10] Ishita: Yeah, thank you so much. And then it was great meeting you as well as having a chat with you. Likewise.

[00:28:16] Luke: Likewise. All right. Excellent. Thank you very much. Thanks for listening to the Brave Technologies Podcast.

[00:28:23] To never miss an episode, make sure you hit follow in your podcast app. If you haven’t already made the switch to the Brave browser, you can download it for free today at brave. com and start using Brave Search, which enables you to search the web privately. Brave also shields you from the ads, trackers, and other creepy stuff following you across the web.