In The Arena by TechArena - AI, Analytics, and Innovation: Insights from Unilever’s Arun Nandi

Episode Date: August 27, 2024

Arun Nandi of Unilever joins host Allyson Klein to discuss AI's role in modern data analytics, the importance of sustainable innovation, and the future of enterprise data architecture.

Welcome to the Tech Arena, featuring authentic discussions between tech's leading innovators and our host, Alison Klein. Now, let's step into the arena. Welcome to the Tech Arena. My name is Alison Klein, and today I am delighted to invite Arun Nandy, Head of Data and Analytics at Unilever to the studio. Welcome, Arun. How are you doing? Thank you for having me. All good this week. Yep. So Arun, why don't you go ahead and

just introduce yourself and a bit about your background in tech and how you've come to head data and analytics at Unilever? Yeah, thanks for that question. I've been in the analytics and AI industry now for close to 17 years. And over this time, I've had a great opportunity to be part of this evolution of what we would call business insights initially. And we've evolved from that into analytics and now, of course, into AI and generative AI. It's been a fascinating space to be in. A lot of learning, a lot of capability development that I've been a part of and had a great opportunity to contribute to a lot of impact for organizations. And one of my personal favorite areas that I've been able to contribute to is just the talent and skin development, the opportunity to hone and build. So that's been personally very satisfying as well.

That's awesome. One thing that I think about when I think about a company like Unilever and just like the complexity of how many businesses you're in. And I think that one of the questions that I have for you is when you are looking across so many different business opportunities, how do you manage data analytics in such an environment? And what is the advent of AI look like when you're trying to pick from use cases in which to employ artificial intelligence? I'd say that the fundamentals of what data and analytics has meant has stayed true over all these years. One of the foundational building blocks for this is a strong and future fit data architecture and data foundation. The shape and size of what that looks like has evolved over time.

In the past, that would be something like a data warehouse and over time that's evolved into a cloud architecture and lake houses, etc. But there's a strong data foundation, which is one of the key foundational building blocks. The second is the ability to answer business questions, which is fundamental, again, in order to drive impact in any large organization. And specifically with ours, as well as others that I'm familiar with of our size and scale, Fortune 100, Fortune 500 companies, focusing on what those use cases are, those functional areas are, and what those business questions are that we're looking to address

is going to be a key element. And I think the third part is this advanced analytics, machine learning, and AI area. This is obviously the one where there's a lot of talk about generative AI in the past 12 to 18 months. We've been working on Gen AI from in the past 12 to 18 months. We've been working on Gen AI from before the term Gen AI got coined. Having been part of that, this is something which is always a fast-moving target and it helps to be abreast of all the technology and the developments out there which are happening in the industry. There's such a wealth of information out there on organizations which

are doing their own fine-tuning. They're doing a lot of model training. And there's so much information out there to collaborate with some of the industry experts and also with the community at large. And these three foundational building blocks, I feel, have been critical. And the meaning of that has changed over time, but they have still been rock solid in their prominence and importance. Now, you know, I was doing research for this podcast, Arun, and you have a lot of notoriety within the AI community as being a top leader in AI. And I'm so glad to have you on the show so I can ask these questions. How has AI adoption evolved since you've been engaged in this just expanding arena?

And what do you see as the opportunity for enterprises today with the tools available and the state of technology? Yeah, thank you for that. I appreciate that. And I think it's been a great journey within AI, as I mentioned. And part of that is a personal endeavor of mine, which has been to be a strong promoter and one that's always forward leaning in the space of AI. And what artificial intelligence and what generative AI has meant to organizations and enterprises in this past two year time frame has changed dramatically. But AI adoption is typically, again, looking at what applied AI means. One of the strongest

contributors to success within that is having a clear determination of what use case you're going to solve. So one of the building blocks that I mentioned. And a framework that I really like to apply, and I've spoken about this a few times in the industry as well, is this inverted pyramid of AI investment, which suggests that 10% of your resources and effort should be spent on the actual algorithms and the models, another 20% on tooling and technology and platforms behind them, and the remainder of the 70% in change management and adoption. Because fundamentally, what I believe is that we're making a process change or a change which affects certain individuals and how they've been going to business in the past several years. And if we have to change that, then the strategy for adoption has always got to be people and

process led. And I think many times industry practitioners confuse that we've got to spend most of our effort on the technology, the tooling and the algorithms. But that's only half the journey done. The adoption journey really begins from there. And I challenge a lot of my internal product teams and a lot of the teams that I work with that we've got to have this strong foundational building block around adoption and what we're doing to really drive the output of these tools within the businesses. I'm glad you brought up tools because it's one thing that I have been studying quite a bit is the advancement of public tools that are available to enterprises. And I guess one thing we see the hyperscalers coming out with their tool sets, we see open source

initiatives with tool sets and some startups that are trying to offer some different solutions to companies. How do you see the confluence of public tools and then your own training and fine-tuning for various use cases working together? And what is the status of industry readiness in this space? Yeah, this is a great question. I think we're at this specific point in time where the advancement of proprietary tooling,

open-source tooling, and enterprise tooling has formed this very interesting trifecta. And specifically in this hyper-revolutionary space that we're in, what I have found useful is to essentially experiment with multiple versions of each within the proprietary model space with the dominant providers out there, within the open source space as well, the dominant providers, and also having a strong build program. Because I always believe that you've got to have a healthy balance between build versus buy, especially in these areas. And I think many companies that I'm aware of, ours included, are really building core competencies, which are investing in long-term competitive advantages by building our own and having a multi-pronged approach and having a strong

decision flow chart for where you build versus buy and where you're buying, where you go proprietary versus open source, I think has a strong ability to help you progress through that journey. I have many variables that I think about in that decision flowchart, everything from intellectual property to the way that data is treated. There's a lot of talk about data residency laws as well. We're in the month when the EU AI Act has gone into effect. So there's that as a variable. So there's many different variables that contribute to how these decisions need to take place. But I think specifically in a time where there is so much evolution happening,

it's important to have your eggs in different baskets. Now, you are speaking at the AI Hardware Summit coming up next month. Tech Arena is a media sponsor there. Why is this such a critical moment for hardware innovation? And what are you hoping to get out of AI hardware? I really enjoyed attending and speaking at the AI hardware summit last year. I think this summit is one of those that has always had a strong pedigree of speakers and attendees. This is where the audience is notably of very high technical caliber and very high intellectual capital as well. And I have benefited quite significantly from many of the thought leaders, Andrew Ng and others that have been speakers on the stage in previous years. And I really look forward to the great speakers and the wonderful agenda that's been put together for this year as well.

And I think this brings together some of the best minds in the industry thinking about how hardware and AI design actually applies towards future facing technologies. those places where if you've got an eye into what you need to do in 2024, but also in 2025, this is an absolute must-have for you to be able to atone and add to your repertoire. You know, we want to get into the hardware a little bit because the hardware vendors will be there next month. It's a great who's who of who are developing innovative solutions in this space. What do you see that encourages you from hardware vendors when it comes to the demands for AI? And what would you like to see the industry do more to address moving into 2025? I think one of the areas that hardware manufacturers have spoken about and one that is important to the industry at large is sustainability.

When we talk about hardware design specifically for AI as well, this is typically linked to compute efficiency. Beneficiaries of this as practitioners who are in this industry, we've obviously moved some way away from maybe 15 months ago when there was quite a constraint on the ability to acquire some of the AI hardware that was out there and there was quite a scarcity of that. But today, even though we are in a land of so-called abundance, I think it's imperative that we invest behind technology that is sustainable. I was reading an article recently about how the top technology, specifically hardware manufacturers and cloud service providers, are incurring electricity usage and wattage, which rivals a few countries.

And I think there's obviously a lot of that that's driven by this AI hardware that we're using. So I think it's important for the industry to move forward in its ability to move the ball forward in terms of not just parameters, accuracies, and the ability for agentic AI to be the next frontier forward, but also how efficient we're able to do this. And I actually remember a great anecdote from one of the speakers at the AIHW summit last year, which said that the human brain actually uses 20 watts of electricity. And just remarkable how amazing this human brain is when you just think about efficiency of electrical output, right? And I think we're some way away from our models being as efficient as that, right? Some of the largest LLMs, as an example, use electricity that rivals probably a few thousand kilograms of greenhouse gases emitted each day. So I think there's more we need to do

in that space to be more sustainable and efficient. I think that's such an interesting topic. And I want to ask you a follow on that. You know, when you think about the compute density required for AI, it feels like we are seeing a re-architecting of every element of the data center from compute to memory architectures to fabrics and then power and cooling technology. Do you think as somebody that's working within an enterprise with so many changes going on right now within compute, how can enterprises stay abreast of all of this innovation happening in parallel? And where are you seeing the most important innovations being addressed by the industry around sustainability? I think there's a lot of work that's already happened and there's more to come in the space of architecture design as it pertains to efficiency. There's two ways to address it. One, of course, from the hardware design angle of how we're going to be more efficient in that.

And I think some of the attendees and the speakers specifically from the hardware side of the equation have spoken both last year and I know will be speaking this year as well on their own evolution in silica deployments in design related to accommodating additional capacity as well and another on the data architecture side slightly downstream is going to be related to designing for efficiency. And I think our products and our solutions and tools today that we're building have to have a solution architecture and a solution design that is able to adopt the most compute efficient and the most sustainable solutions. One of the ways in which this is done, thinking about the world of LLMs and all the use cases that we've got running in those is through something as simple as query routing. So you have an ability to route queries into different kinds of models. You perhaps do not need your most complex multiple billion parameter model to be

activated and called with an API every single query. You might be able to address some of those through non-LLM calls. You might be able to address some of those through non-LLM calls. You might be able to address some of those through the smaller parameter engines, and you might want to leave some of the multi-step complex queries into the more advanced model. So there was a lot of talk about small language models in the last year's summit as well. And I think that is going to continue to be one of the themes, which is more domain specific, more function specific models, which are quite a bit more accurate and have the ability to be tailored to specific domains. There's a lot within solution architecture and design, I think, which can go far to help us as an industry move forward towards sustainability.

That's fantastic. I can't wait to hear your session at AI Hardware Summit. And I'm sure that folks who are listening online want to engage with you and continue this conversation as well. Where can folks reach out to you and connect to continue the dialogue? Yeah, so I'd encourage people to reach out to me on LinkedIn. I am quite active as a practitioner in ensuring that we as an industry community come together. In many cases, I've had the great opportunity to participate in public-private enterprise academic collaborations. And I think it's a great step in the right direction for us all to come together. In many cases, enterprises are dealing with very similar challenges. And I think it's not required for everybody to reinvent the wheel.

There's a lot of learning and sharing that the community can do, which can really help us. So fully open to that and happy to have people reach out to me on LinkedIn. Fantastic, Arun. Thanks for the time today. I would love to continue the conversation on the tech arena as well. Thank you,un. Thanks for the time today. I would love to continue the conversation on the tech arena as well. Thank you, Alison. Thanks for having me. Thanks for joining the tech arena. Subscribe and engage at our website,

thetecharena.net. All content is copyright by the tech arena. Arena.