#185 Data platforms: the foundation of business-oriented data analytics with Aruna Kolluru, Chief Technologist, AI at Dell Technologies

---

In this episode, Felipe sat down with Aruna Kolluru, Chief Technologist for AI at Dell Technologies ahead of our Advancing AI Melbourne event on 6-7 April. She shared how they work on providing solutions for their customers and use all available technologies - AI, IoT, data, etc, - to reach their desired outcomes.

Aruna works with clients across a vast range of industries and this is a testament to the power data & AI hold to promote growth and generate business value, regardless of the area you work in. In her own words, data has become the core of innovation for every industry.

With the key role data plays in today’s organisations, making it accessible for the analytics team to extract insights from it is of the utmost importance. That’s why data platforms have become essential for providing reliable, quality data that can be leveraged to achieve business goals.

Some common challenges Aruna has identified are most data and analytics projects are viewed as technical initiatives without any alignment to the business objectives, and the difficulty business users have in finding and using data that exists in a variety of different formats and locations without having a catalogue or a way to really see all the data they have. She fears some business leaders may view data platforms as just another technology improvement project, when in fact they can be the foundation that ensures data and analytics are set up to support their business goals and produce meaningful insights from all the available data. Once organisations have set up a data strategy that deals with the lineage of the data, its quality and the whole governance around it, it’s easier for the data scientists to analyse it.

Other important factors that can help level up your analytics efforts and the impact they have, according to Aruna, are investing in the right analytical tools and fostering a culture of constant learning within your company that encourages employees to improve their skill sets and keep up with the industry’s rapid pace of advancement.

Aruna also shared some of the basic steps to follow when it comes to end-to-end delivery:

1. Identify the right use cases.

A tip she offers is using examples from your peers of what they're doing or taking inspiration from other industries. You can also pick your most pressing challenges and assess how AI can be leveraged to solve them, but doing AI just because it's ‘the coolest technology’ is never the way to go.

2. Recognise gaps in capabilities and prioritise what you want to achieve.

Think about the feasibility of the solutions and assess if you have the necessary data and capabilities.

3. Do a proof of concept.

This will bring out any challenges that may arise in real projects and help you prepare to avoid/ overcome them.

4. Strive to work within a scalable and adaptable infrastructure.

Technology will change, the way we analyse data and algorithms will change, so think of how you can avoid problems like data migration in the future, by choosing a platform that can scale and adapt with your needs.

Tune in for the full conversation with Aruna on how to leverage data platforms. You can also join her in person when she presents with David Siroky, Asia Pacific Head of AI, Data Analytics & HPC, Director, Dell Technologies at Advancing AI Melbourne (6-7 April, 2022).

Huge thanks to our Diamond Sponsors, Dell Technologies and Microsoft for supporting this event for the data community - to learn more from them register here:

Resources

Learn more about all the ways Dell Technologies can help you leverage your data!

“Data is the key. And there are quite a few challenges as data is not something you can go and buy from the market or it’s your data, and you’ve been collecting it for years together, how do you actually make it accessible and bring out those outcomes from your data? That’s the biggest challenge”

— Aruna Kolluru, Chief Technologist, AI at Dell Technologies

What We Discussed

Are there any particular industries that you work more in or less than?

The role of the data platforms in organisations today, how are you seeing that being experienced throughout different industries? And what impact do you see that happening out there having in businesses today?

Are there ways that you see that people are improving their data platforms in different industries, or strengthening the process to create anything around improving the reliability?

What do you see as the best practices on putting those components into the work to deliver initiatives in organisations?

What does a good POC look like on the data platform side?

To a problem of trust of the quality and the preparation of the data that is available in the catalogue. How have you seen organizations? Or how do you advise organizations, to tackle that part?

What is a good POC look like from your perspective?

What do you propose to executives and leaders that are navigating these, treacherous waters at the moment when it comes to talent and building teams?

What type of use cases do you see coming up in that space?

Do you think that their applications were said ready to productionise?

Are there any areas that you're particularly passionate about? Any anything that just really gets you excited within AI?

What do you see as the ways that people can get into federated learning, and some of the challenges that they should be prepared for?

Graph technologies- What do you see that people should be thinking about applying that technology? And where should that technology be applied?

Episode Highlights

We don't see each solution as a separate silo, like AI, IoT, or data or anything. If we're going to get to an outcome, you need all these technologies together. We focus on all of them.

That's one of the challenging and interesting parts of being in this role. We are talking to an agricultural customer in the morning, and then maybe a different one in the afternoon, and then a financial organization. So it's a range of things. And what we see is, technology can solve multiple problems we learn from different industries. At the same time, you see these use cases which can cross-pollinate from one industry to other as well.

In my experience, most data and analytics projects are viewed as technical initiatives, without any alignment to the business objectives. That's where they fail to align data and analytics to business priorities. Business leaders must consider them more than just like our technology improvement projects. This will really ensure that data and analytics are set up to really solve and support their business goals and guarantees that they'll be able to produce meaningful insights from the data.

Do not do AI just because it's the coolest technology, it's always important to have business buy-in and involvement, while identifying the use cases. I'm not sure of the order, but we need to consider consolidating, or even say, having a catalogue of your data, so data scientists know what data is available for analysis. And then what I always say is, it's important to have a data platform where you don't need to move your data, right. So your data keeps on growing and data migration is the biggest challenge. So you need to have a way where you don't have to migrate your data, have a platform where you can just put in your data catalogue so it's available for access all the time.

Do POCs, it's important to do a proof of concept, it will give you a flavour of success as well as challenges you might face in real projects, whether it's in terms of the data or skills, or it might not work, right, you might actually go through that process of iteration again, and again

Doing a POC is really good for identifying some of those challenges and working on them. And then start small, don't boil the ocean, there are plenty of things you can do with AI. Pick something you get most value with less effort

We need to think about solving a simple problem. Right? Don't really think like I want to make the most of this POC, but like a simple problem where it actually shows business value. Have the business mind tell them like what do you need. And it's important to have a team of people here I've data science is a team sport, we need to get the business people, the data scientists, the data engineers, all of them together. This particular POC is not just about technology, but it also tells you about what kind of team you need, what kind of diverse teams you need there to build that success.

This reminds me of a study I recently read in McKinsey, that high performers are 400% more productive than the average ones. And in complex of occupations, like data science is one of them. high performers are 800% more productive than the average ones. Wow. So actually see all these AI occupations, right, whether it's data scientists, data engineers, all of them fall in this complex occupation list, there's a huge demand and they're hard to retain, for multiple reasons, right? Like one is like you said, there's a lot of movement, there's a lot of requirement, but at the same time, from an organisational perspective, I would say build your own data teams within your organization. I see a lot of organizations hiring contractors, but it's highly recommended to build an in-house team, because they are the ones who understand your domain, they are the ones who understand your business.

Look for individuals who are curious about learning who are passionate about learning. These technologies change every day, so we want someone who's curious and passionate, rather than someone who's experienced in that, right. So it's a combination, of experienced people as well in the team, but it doesn't have to be a complete PhD. Of course, they actually need to have a solid foundation on the basics, but it's good to actually get these people who are passionate and ready to learn quickly on the changing trends and the new technologies all the time.

So tiny ML is, again, machine learning, which can be deployed onto microcontrollers, right, very small devices. And when I was thinking about that, that can actually run battery power, it doesn't need any network connectivity and it can actually run for days without any power or network or anything of that sort. And I was thinking like, how can we actually embed this somewhere where we don't have any of these resources, that's when I thought, we have these bushfires all the time in Australia, and if we can have some of these in the forest, and then they keep monitoring it, maybe just sound an alarm or something, or they just keep pinging when there is network connectivity, a little bit, and if multiple of them are going wrong, there's something wrong, right? We can actually expand it right, it can go with a camera, and it can watch and it can say yes, there is a fire when it is a really small fire or smoke or whatever it is, or even conservation kind of thing when it is small, and then send that image or picture or whatever to the even the video footage of it when there is a small fire so the fire troops can go there and stop it before it becomes big.

And then we also have pre-trained algorithms, right, for something simple. Think about crowd counting, counting the number of people, there are plenty of use cases, you can use it to really see how a particular road is being used, or how a particular building is being used, whether it's in schools, or shopping malls, or airports, or how long the queue is, how can we actually support like we provide services for the number of people waiting in the queue? How many people do you need to serve the people into that number of people, the plenty of use cases, but again, this particular algorithm is already there.

In a traditional world? It's hard to really find those connections, but the graphs? It really, it really opens up that whole new world of finding connections, whether it's for recommendation engines, finding fraud in transactions or anti-money laundering, all these different use cases, open up with graphs.

At Data Futurology, we are always working to bring you use cases, new approaches and everything related to the most relevant topics in data science to help you get the most value out of these technologies! Check out our upcoming events for more amazing content.