Chris Royles of Cloudera on curiosity, data modernisation, and the future of AI

Todd: Hello, and welcome to a very special extra episode of CWO.digital. Today’s guest is Chris Royles, CTO for EMEA at data specialists Cloudera.

With a storied career that encompasses everything from computer graphics to robotics, and some of the earliest ever applications of artificial intelligence, Chris is a technologist with a passion for powering change; and there seems to be no limit to what can be achieved with the right combination of data and ambition.

And so without further ado, I’d like to welcome Chris to CWO.digital!

Thank you so much again for joining me on the show. Why don’t we start by getting to know you a little bit better? Could you let our listeners know what path led you to the career that you have now?

Chris: Certainly Todd. I’ve been with Cloudera for what’s coming up to a decade in August. But if I go all the way back to the start, I remember when I was at school going through my A-levels, uncertain of what I wanted to do, and I had this aspiration of building things. For some reason I thought architecture felt like a good place to start. I’d get to build big things. But by fortune or misfortune, I decided to take a different path, and part of it was down to me very simply not being able to spell the word architecture…

So when I was asked at a jobs fair what I wanted to do, I wrote down computer science, partly because my father was doing computing at the time. He was helping run steel and manufacturing processes using computing, so I was introduced to computers at a very early age.

I went to university, studied computer science. I found myself naturally drawn to things like computer graphics; how they function, how they operate and really experimenting with that sort of technology. Everything from games to building three dimensional objects that you could manufacture, design, and produce. I really enjoyed that, and it led to me carrying on to do post-doc research.

After I’d done my degree, I stayed on and did a PhD. The reason for doing that was more by accident. I saw they were recruiting at Liverpool University for doctorates in artificial intelligence. I thought, ‘that sounds really interesting. Don’t really know a lot about artificial intelligence’. Spending four years researching it might be quite useful.

So just before 2000, I finished my PhD in artificial intelligence. Back then it was very focused on expert systems; less on neural networks as it is today. Ironically, back then AI was very predictable. It would give great answers that you could then check and validate. From that, I then started a career as a software engineer, designing telecommunications, doing voice analytics and voice response. Then I transitioned into the sales function of businesses through a company called MapInfo, working with geospatial mapping and data. That’s what brought me on my journey to field CTO at Cloudera, where I really care about technology and look to help our customers gain the best they can from it.

Todd: I’m really impressed by that sense of curiosity you seem to have; spotting things that you don’t know about and letting that guide you. I think that’s really admirable.

Chris: What I’ve found is that as my role has developed, I’ve had options. Do I want to carry on in an interactive type of role where I can speak to people and do practical things? Or do I want to transition into management and manage people and manage teams? What I tended to do is transition toward what we call an individual contributor role.

I tend to mix my time between two key things. One, talking to people, mostly customers. Asking, what are your problems? What are your pains? What are your challenges? But second is our product function; how can we fix, how can we help, how can we design? I try and focus on how things get built, the engineering aspect of how things work.

Down to fundamentals, I get drawn into various topics like security or accessibility. Details that are really important in technology but are sometimes overlooked. My finding is that people don’t necessarily always understand why they’re important. So it becomes part of my role to explain that, not just to product and engineering within their own business, but also bringing that insight and understanding to our customer base as well.

Todd: When it comes to data, there are so many variables. It can be really easy to become distracted. When you are taking part in a larger project, how do you maintain the big picture view that’s needed?

Chris: Great question! It’s not easy. I always use the phrase, ‘let’s step back and look at the bigger picture’. We get drawn very easily into listening for the first pain and then trying to solve for that, rather than listening to the pain, clicking into the next level of detail, then clicking into the next level—to the root cause of why that pain exists. What you find is sometimes your first instincts are not necessarily correct because there might be a broader issue that needs to be addressed.

I also like to reflect on technologies. How people and process come together; especially in a world where there’s more automation. A good example we are being asked a lot is around artificial intelligence and the governance of it. It falls into three categories; first, is my data ready for use in artificial intelligence? If it is, how do I make sure I bring that data together in the right way and follow its lifecycle all the way through to use? And then when I use it to create a model or an agent in a network of collaborative agents, how do I make sure they all perform well individually and as a group?

It’s about stepping back, looking at the bigger picture. I tend to ask my teams about the current state. Then we can explore that in detail. Then what would the future state be? How do we think of that as a bigger problem to solve?

Todd: From your experience, is there one challenge in particular that you keep seeing people struggling with, time and time again?

Chris: In the business I represent now, we’ve focused on helping organisations bring together very large amounts of data. And when you think about large amounts of data, that’s become relatively commoditised in terms of how technology can solve that. Storing large amounts of data is no longer a really hard problem. The harder problem is how we get that data into a form that’s useful to a person that can use it.

What I’m finding is a lot of the tools I’m using today in artificial intelligence don’t necessarily make me more efficient. I have to work quite hard to get them into a form where they work for me. What’s important is making sure the technology gets out the way rather than gets in my way.

My advice really is to focus on how you get the data to the people that need it, and how you then craft it in a form where it really is truly useful to them.

Todd: Great advice. That element of usability does seem to fall by the wayside quite easily, when it should be at the core of absolutely everything.

In a broader sense, how can Cloudera help our listeners achieve these goals?

Chris: When I went recently to a number of events around data and analytics, it’s like every event now has got some dimension around artificial intelligence. What I enjoy doing at those events is meeting customers at different levels of where they are in terms of their problems. In an environment where you are forced to talk very concisely and very quickly, you can really get to the point quickly.

One topic we focus on is, how do I get my data AI-ready and that trends toward the process from source to data management. You might refer to it as the lineage of data, how it gets to where it can become useful.

The second piece is how we get it useful, but in a private way. We refer to that as private AI. The reason for both these is that the most important bit in the middle is the data. It has a gravity. People come to data because they think it’s got use and value, and it’s easy to bring tooling to data. It’s far harder to move data around.

Something like a large language model has got large in the name; but they’re very portable. You can download them off the web. It’s a lot harder to download 10 petabytes of data, especially when that data has to be under management and secured.

We really want to help organisations choose where to bring their data under management. If it’s born in the cloud, that might be a great place to store and manage it. But if it’s already in the data centre; say in a regulated industry like financial services, then shouldn’t the models come to the data?

Shouldn’t I be able to run my models on data that’s born in the cloud and treat my workloads as portable between those environments? Then when you think in terms of something like a model that’s trained and built on data, you’ve got to think in terms of what that constitutes, including what could be personal and private information. It also includes a degree of your own intellectual property as an organisation. It’s an asset in its own right, and it might be differentiated for you and your marketplace and the types of businesses you are trying to operate.

So we’re thinking about the protection of data, but also the protection of intellectual property; where that sits and runs and where that’s operated and by whom. Those are things I think I’ve heard most from people I’ve met at these events.

Todd: It’s such a crucial intersection of these different things, all coming together to create a really knotty puzzle. That’s great that Cloudera are there to lend a hand. It does make me wonder… Ten years from now, what do you think will be the main difference in the way you work if things continue to change at their current pace?

Chris: Ten years is a very big horizon. When I think about horizon scanning, I look at five years and maybe even one to two years. In terms of the trends I’m seeing in artificial intelligence, things are moving so fast. Everything’s changing all of the time. So if everything’s changing all the time, your core systems are evolving. The data you have in management is evolving and changing the applications that your end users are accessing.

We have a concept within the platform called the Lakehouse. Financial services is the best example of this. In financial services, a bank will have to generate multiple regulatory reports. What they would do before is take the data, then create a copy of that data for the report, then send the report out with the Lakehouse. What it means is they can effectively create a version or a branch of that data and then share that data without copy.

The regulators, when you speak to them, their challenge is slightly different. They get hundreds, if not thousands of inputs coming from many different organisations, and somebody breaks something at some point, they’re passing data that’s got the wrong structure, the wrong schema, the wrong shape, or changes in frequency. So they need to be able to change their schemas. They need to be able to evolve them. They need to be able to change the rate at which data can be processed and they need to be able to change the partitioning of that data so it’s efficient when it’s analysed.

The Lakehouse just kind of makes that transparent to the platform itself. You flow the data in, many of these things will be automatically accommodated. It’s a very simple example of how this Lakehouse approach; where you are versioning and branching data, can be used to move an industry toward achieving constant change. The pipelines will effectively self-heal. We’re deleting complexity in the process, and I think that’s kind of interesting. Things like the Lakehouse are layering up and enabling organisations to do that with fundamental data management, but also bring different workloads to that data so they don’t have to create lots of different copies of it.

This is really driving a degree of transformation in how organisations are not just managing their data, but thinking about how their organisation changes pace to becoming a type of business where everything could change all of the time.

Todd: This is a question I ask every guest we have on the show. When you and your peers all get together, what’s the exciting thing that you are all keen to talk to each other about, that the rest of us don’t know about yet?

Chris: It can vary dramatically. My peers will each have different areas where they’ll specialise. Or if I’m engaged with a client and talking to peers in my customer base, they’ll have particular focuses around what they’re doing within their business. Either transforming their business, or maybe taking costs out of the business. The most common aspect I find when talking to technologists is wanting to simplify and consolidate. We hear that a lot. I’ve got 400 applications, and I want to reduce that to 20. That might be a good example of the types of conversations I hear quite often. What I also find when I’m able to have more open discussion is about what the future holds. Where do you think the future’s going?

Artificial intelligence is already well understood. People are starting to use it. It’s heading towards the plateau of being useful and productive. You see these things progressing and then you see things that surprise you, like agents; how are agents going to be used in organisations? I think some of the analysts and the research I see are putting agent networks out three to five years before they start to become useful. Organisations that are adopting early are already starting to see value, and I think that’s a really interesting dynamic. We’re already starting to see technology that’s seen as bleeding edge as being highly productive.

There are certain techniques that you kind of get for free that people overlook. I’ll give you an example. For some organisations, they will interact with customers across multiple languages and regions. Yet when they think about large language models, they forget the basics around being multilingual and multimodal. These models can do voice and transcription. You don’t need to have lots of different models to do that. So they miss what you get for free.

Another example is that models that do voice analysis can be very low cost to run but have a very high return on their investment. They can really significantly reduce the time it takes a call centre handler to handle a call. It can reduce call times by three to five minutes. It’s cheaper to run, but it has a higher return in terms of time savings and efficiency. That disproportionate offset I think is really interesting, and people don’t look for that as much as they should.

Todd: That is interesting. Thank you very much, Chris. Unfortunately we’re nearly up on time for today. But before we go, could you recommend any books or documentaries that touch on some of the topics we’ve chatted about?

Chris: I’d certainly direct you toward the AI Forecast. It’s on Spotify right now and is sponsored by Cloudera. It brings in a range of different experts in the industry and gets into the detail of artificial intelligence.

I don’t necessarily have recommendations for books. I very rarely read books cover to cover. I will dive into a particular topic, research it in some level of detail, and then what I prefer to do is get hands-on. Within the Cloudera platform are AI workbenches. Whenever we release something, I log in and I just get on working with things. So I actually spend much of my time questioning copilots on how to optimise source code. I’m using AI much more in my research processes. I like it to be in context of the application, so I’ve been using the copilot embedded within our own development tools; not only to help me write better code, but also to research documentation. Inquiring around very specific topics to move me forward, step by step and incrementally on particular topics of my interest that I’m trying to solve. So I’d recommend people take that approach, rather than go and read a book. I’d say have a conversation with a large language model and do some deeper research around a topic that’s most of interest to you.

Todd: You heard it here first listeners. Read fewer books! That’s Chris’s official advice (laughs).

Chris: I won’t say that to my children! My advice to my children is read lots of books because there’s nothing quite like picking them up. But I don’t tend to sit through a whole book anymore. I will dip in and out, but much of that information is available online and already in the corpus of the large language models you might use every day.

Within Cloudera, we use AI extensively. We have access to a number of different services as employees of the business. And I use that very extensively for my own research.

Todd: Thank you so much, Chris. You’ve been an excellent guest. Thank you for joining us.

Chris: Todd, it’s been a pleasure speaking with you today. Thank you.

Todd: I’d like to thank Chris Royles again for joining me on the show, and thanks to you for listening. You can find more episodes at CWO.digital, or wherever you get your podcasts. Until next time, thanks again.