Hypercube Podcast Transcript

Title: Delivering Successful Data Initiatives in the Energy Sector

Host: Adam Sroka
Guest: Rosina Norton

Intro: Welcome to the Hypercube podcast, where we explore the cutting edge of data and AI in the energy and utility sectors. I’m your host, Adam Sroka, founder of Hypercube, a strategic technology consultancy supporting asset owner operators, traders, route to market providers, and energy services companies unlock the power of data and AI.

Join me as I sit down with experts from across the energy industry to uncover how they are navigating today’s data driven challenges. And without further ado, let’s get into today’s episode.

Adam Sroka: Hello and welcome back to the Hypercube podcast. Today I’m delighted to be joined by Rosina Norton. Rosina, would you like to introduce yourself? 

Rosina Norton: Hello, Adam. Thanks for having me on first and yeah, to introduce myself, I’m an independent data consultant who specialises in clean tech slash climate tech, because I think they overlap a lot, industries.

So right now I work with a range of different companies to help them understand how they can better use their data to bring value to their business. But I come into that from having 10 years in the. clean energy industry, always in data roles ranging from being an analyst as I first started my career to more recently finishing off as head of data science.

So I’ve had a lot of exposure to that kind of journey through the analytics stack. 

Adam Sroka: Very interesting and it shouldn’t come as a surprise to the audience that I absolutely love talking to other consultants. I think that we’re quite lucky that we get to see quite a broad range of organizations and businesses and get like unique perspectives that others don’t.

Do you find that kind of holds true? 

Rosina Norton: Yeah, 100%. And that was actually one of my kind of motivations to go independent was that I’d just finished off a tenor of being head of data scientists. In a big organization, which had grown a lot since I’d been there. And I was interested to think, Hey, what do I want to do next?

Knowing I wanted to stick in the clean energy sector. And then luckily it was in a place I already had a couple of people who’d reached out to me about independent work. And for me, it was like a win I get to both learn. Tons about all the different areas of clean energy and clean tech at the same time as working and bringing value to a ton of really cool organizations.

I really love it. 

Adam Sroka: So you’ve obviously seen a lot of organizations and problem sets and you’ve got quite an interesting experience, what’s your view on the biggest challenges and opportunities that the energy sector is facing with regards to like data and AI? 

Rosina Norton: There’s a lot going on in a lot of spaces.

Which I think makes identifying like the biggest challenge and not the most easiest thing to do. But I think what I could talk to is like the opportunity and what excites me. That’s happening right now in the industry. For example, I think one thing, that I’m noticing a lot and particularly through my consulting is the kind of merging of the more traditional engineering firms and ways of working and more of the tech focused ways of working.

I think there’s a lot of engineering firms, project development firms, which are really strong on that engineering and kind of operations point of view of like, how do I actually develop this solar project? And have access to some of the kind of incumbent tools that might be used for that to available to GE and other big engineering firms like that.

But what we, what I’m seeing more and more is some of those companies getting more interested in the tech way and the more kind of modern data stack as I could put it, of thinking I have all this data and I hear I can do really good stuff with it. But right now, all my data might be stuck in CSV files and I’m using it in MATLAB.

But actually, how can I make that more flexible? What kind of data infrastructure could I build that could host that data and make it available for other purposes? And what could those other purposes be? And how do I get access to them? So I think that’s maybe one thing that is, And maybe a bit of a change that I’m noticing over the last couple of years is really that kind of merging of maybe you could call it like the old school engineering world with the more kind of modern tech world.

And I’m seeing quite interesting and exciting results coming out of that. So that’s one thing that I think’s really exciting. Other opportunities as well. I’ll be honest, a lot of my focus at the moment is in the kind of global South, particularly sub Saharan Africa, given that’s a lot of where my career has been focused up to this point.

So I see a lot of opportunities there, which might not be directly applicable to Europe or the U. S. because it’s a very different economy and lots of different dynamics going on. But for example, one of an area which I think is really interesting is resource, kind of resource measurement and estimation.

It takes me back to the wind days where we would be measuring wind on site and then having to correlate it with long term references. One thing that I’ve discovered working in some of these developing economies from A clean energy perspective is you don’t have as much readily available public resource data for wind or solar.

So like in the UK, you can get hold of the metrological office data and they have masts all around the country. So you can actually get quite high granular data points for long term solar or wind resources across the country. That just doesn’t really exist in a lot of these economies you’re working in.

And you’ll reliant on more high level data sets, such as like satellite analysis data that you can’t get from nowhere and places like that, but it’s nowhere near as accurate or gives as much granularity as you would get from local resource estimation. So that’s one challenge I see is that is a wider problem is that obviously all renewables are intermittent.

Natural resources, which as part of them are a variable resource and to understand the kind of dynamics of your project, you need to understand the dynamics of that variability and to understand that variability, you need the data. So I’m excited to see how that’s changing in some of these regions that I work a bit more in so you can get a bit more certainty into our kind of project analysis early on.

Adam Sroka: That’s quite funny because I’ve worked with plenty of people in the UK that can’t seem to get hold of that data over here, even though it is available and it’s there. But because that’s the thing. We take for granted the data sources that we have available to us, and people struggle with just the integration of the use of them, but yeah, what do you do when you’re in an environment where that doesn’t exist?

That can pose a significant issue for if you’re trying to make investment cases and build big projects of this kind. 

Rosina Norton: Yeah, I think that’s something. From a kind of clean energy perspective. And then another area that I think is going to be applicable in the industry, but it hasn’t quite got there yet, maybe, but there’s this idea of like tiny ML in that right now, a lot of the machine learning algorithms we use, they’re having the success they have because we can throw a lot of compute power at them and a lot of memory.

And they usually take up a lot of space locally, if you’ve got it stored on a device or something like that. But I love the kind of TinyML community, which is looking at how can we get performant machine learning models, which work on a much smaller footprint. And typically they’re used for like edge devices or against is where I’ve come across them in areas where you’ve got low connectivity.

So you might not want to have your ML model hosted up on the cloud and depend on a fragile kind of 3D or 4D connection to get that. You might actually want to have it on your device where you can. And then collect any recordings or predictions in batch from there. But I think that’s an area which is going to be really interesting for the energy industry because we know that the kind of intersection of energy and AI right now is under a bit of a microscope because of the energy intensity we’re seeing have come some of these big LLMs and big AI models and with the EU.

AI regulation coming through and companies going to have to start potentially reporting on energy intensity of this right now generalized AI cases. It does make me think, is there going to be a bit of shift in consciousness about the resource demand of all these kind of advanced AI and ML models that we’re going to be using and Are we going to push more towards these kind of like tiny ML algorithms and especially when you’re thinking about predictive maintenance that you might have on an operational site?

Would you rather have that sitting in the cloud or you could have that on site if you had a really efficient ML algorithm? I think that’s another area which really gets me excited. 

Adam Sroka: Yeah, and I had exactly that issue a few years ago for a customer where, yeah, there was a really tight requirement on the latency.

Other round trip latency of kind of signal to prediction or decision back to device. And yeah, the only solution that would meet all the requirements had to be deployed on the edge. And it was quite interesting to do that a few years ago. And it’s good to see it coming more and more to the front. And I think, yeah, if you want to get really fast, really responsive technology.

It’s certainly a way to go, although it just comes with its own headaches and challenges. It’s not the kind of easiest solution to deploy. Just change gears a little bit. We’ve talked about a few things in your career path, but I’m interested. Yeah. What just like your day to day look like if you’re working with a customer and what sort of tools and technologies are you using mostly or in most favor of?

Rosina Norton: Yeah, so day to day, have a good balance of keeping a good chunk of my time for my existing clients. So on those days, we’re working on pushing forward the project we have together. And then obviously keep a couple of days as well for my own kind of business admin and networking, which I really enjoy kind of meeting all the cool people who are working in clean tech and climate.

That’s another really big positive of working in this space is you often meet people who are really like minded, really motivated, and I just generally find it really dynamic and inspiring community to work in, and also doing a lot of research, like one thing that I love about working in data, working in energy, is An old industry, it’s very new and it’s changing a lot.

So keeping up to date with the latest trends, reading like analysis of different, I dunno, a new policy that might have been released or a new tool that’s been, that the data community is talking about. I like to keep a bit of time aside to make sure that. And keeping up to date with all those kinds of things.

And then in the contracts I work in, I have found that I plug most naturally into kind of small medium enterprises, which going through a bit of a transition in their data team, in that they might have a piece of software already, that’s like the transactional system, but they haven’t really put much thought into the analytics aspect of that.

So how do they go about taking some of their. OLTP system data and putting it into an OLAP system so that then it can, and what can they then do with it either as products to drive back into their technology stack or as like an internal product for investors or for decision making.

Either that kind of software tech profile or companies, which are, as I mentioned earlier, in that, Hey, we are an engineering firm. We develop projects. We manage projects, but we don’t have a really an internal data team yet, but that’s something that we want. How do we go about building that? What’s our foundational?

technology going to look like. So the kind of pattern that I’ve found works quite well is starting off with that data strategy aspect, plugging in and learning about their problems, what they want to achieve, what are their goals and helping them put together a data strategy and a roadmap.

And then if they want to continue working together from a technical perspective, where we build POCs. And actually start implementing some of that, and then on the alongside of that, the kind of team development, so supporting them with hiring that foundational hire that’s going to drive their new department, or working closely with the existing IT team or software development team to help them understand a bit more about the data aspect of things.

So it’s really hands on. I like to really understand the business problem. I try not to have too many contracts going on at the same time, because I feel that business context is so important. Domain knowledge is so important. And I don’t feel like I would have the time or space to really get that depth of understanding if I had too many contracts running on it at once.

So I typically have two to three maps running at the same time. 

Adam Sroka: Yeah, that makes a lot of sense. And I think the. Domain expertise makes all the difference from a, like a generalist provider of consultancy or tech services where you really understand and can speak the same language as people. We found that at Hypercube that just being on the same page and being interested in the same stuff makes a big difference.

Do you find you have to do a lot of education for the need for a data strategy or do people get it early on? 

Rosina Norton: Yeah, I’ve been quite lucky that so far it, I’ve had. People coming to me being like, we want to do this. Can you help us do it? Rather than having to push that agenda to say, Hey, I think you need to adopt a data strategy.

It’s more being the opposite of people aware of the potential of data, understanding that they have some useful data, but not. sure how to connect those dots and needing someone who can come in and talk both like the business language plus the how to build a team around that plus the technical to get them on that journey and send them off in the good direction.

Adam Sroka: And what about like expectation management? Because often you’ll find people that like make the decision they want to do this stuff and they want everything yesterday. Is that quite common? 

Rosina Norton: Yeah, I think that perennial problem across all probably industries is everybody wants everything yesterday.

But I think one thing I’ve really, I tried to practice myself as well, even when like internally, when I was a data scientist, is really trying to focus on like incremental value. I think data is, as a data analyst or a data scientist, it’s your job to be providing insight to be. Giving people information to help them make better decisions.

So I think it’s a profession which is really susceptible to like trust degradation. If you get it when, if you get it wrong one, that’s okay. But if you get it wrong twice, once that like kernel of doubt is implanted in someone’s mind, I think, yeah, you can really suffer from a lack of trust, even if you manage to bring things back up to high quality.

So I think. Even for myself and my kind of like own practice when I do my work, I would really try to narrow down the business problem and focus on one or two objectives if I can, and focus on incremental value so that both you can keep momentum and buy in, but also provide rapid feedback to pivot.

Because I think when you’re starting from zero or low, when you’re trying to build a data team, Costs can rack up quite quickly, like the technology costs a lot, not a lot, but It’s a significant cost. You’ve got the people costs as well, that you want to make sure that you’re getting ROI basically.

So making, that’s one thing I often stress to like clients is that and I’ll be honest with them and I’m sure we’ve all read the reports. I think there’s that like famous Gartner report, which is like 80 percent of data initiatives fail. It’s like getting this right isn’t easy. Like you have to stay focused.

We should focus, build a good foundation. And then focus on one or two value propositions and work towards them intuitively, because it’s so easy to get distracted by all the other things you could be doing. And I think as data professionals as well, the work requires very you need deep focus time.

If you’re trying to solve that problem or build that data pipeline. And I think if you have too many things going on at once, you’ve got the opportunity cost of all that context switching. So I really encourage people to narrow it in and iteratively deliver value to try and manage those expectations rather than diving into work where you disappear for three months.

And then you come out the other end with something that might not meet expectations, like making sure you’re really iteratively delivering. 

Adam Sroka: Love that. Yeah, I always say to people what’s the smallest thing we can do to take it To get a little bit more learning and a little bit more understanding and try and build up to a complicated big solution.

That way you mentioned costs of technology costs and data platform costs. Have you got any tips or tricks? Some kind of setting expectations there or how to optimize the kind of long term business case? for investing in a data platform and the technology. 

Rosina Norton: Yeah. Yeah. I think about costs a lot. I think having been head of data science and then also though as an independent consultant where you’re constantly pricing projects or pricing time, it’s something that’s often at the top of my mind.

And I think one approach that I. Take with the projects is again, it’s like data is power. And in this case, it’s not the data that you’re working with. It’s the data about how much the system costs. So I really love platforms that give really high granular. Cost breakdowns, rather than other platforms, which might try to make pricing look simpler by abstracting it out into different kind of units or compute what units, but when you don’t know what that links to, I’m like, that doesn’t help me.

And I actually was thinking about this the other day and it made me reflect on a situation where in one of my previous roles, like a lot of data teams, Compute costs were getting quite high. We were trying to figure out, okay, how do we reduce this? And all it took was like a day or two of analysis of our costs.

Thanks to a really granular cost management platform for us to identify that so much of our costs was coming from the retraining of one specific ML model that we had in production. And we were able to, again, just in a couple of days, make a few tweaks to how we trained the algorithm. And also understand a bit of the trade offs between.

Training, retraining frequency and performance and adjust our retraining schedule and just absolutely slash those costs. And I think for me, that was a really strong learning point as a leader in data that again, yeah, just having that granularity is so important. And I think it is something that we need to be really conscious of because.

We hear a lot about cloud compute costs ballooning and data teams not being returned like ROI positive. So I think as someone who wants to take responsibility for that and make sure that people are getting really good value, yeah, just try to really understand your cost generators. 

Adam Sroka: That’s interesting.

And I guess. It’s one of those things, thinking back to like skill sets and backgrounds that I find a lot of people struggle with, actually, especially data scientists. It’s like the, some of the cloud platforms can be quite opaque or quite intimidating if you’re not familiar with them. And just getting an understanding of how to dig about I’m thinking of the big, the major cloud platforms here, but getting familiar with like how you dig about and find out like, what resource links to what cost and what subscription and how that trends over time.

I think it’s actually a really useful piece of like self development that you can do to start to understand, like the kind of cost benefit trade off of the things you do and that exactly. Yeah. Okay. We want to trade our model every day. Great. But it’s turned out that it’s got a huge overhead and it costs us loads and are there tweaks we can make to optimize that or do this?

So having that pragmatic kind of zoomed approach. Yeah. 

Rosina Norton: It also pushes you to understand what’s happening under the hood. I had a interesting use case recently where I don’t want to name names, but we were using an ETL tool and we wanted to use the low code kind of interface they used for transformations.

And then I said, I implemented them and then checked our costs and they’d like, yeah, they’d probably tripled. And I was like, Oh my gosh, why is that so expensive? These are tiny datasets. This is for POC and kind of pushes you to look actually what’s happening under the hood. And that low code tool, just as default, it uses Spark clusters.

Under the hood to apply the transformations. I was like, we’re working with a tiny dataset here. There’s no way we need to use Spark here. So I think it can also. It’s a useful exercise to rather than you just blindly using the functionality out of the box, which is available on some of these cloud based tools, actually being a bit more critical and thinking what’s actually happening under the hood is actually what I need for the project I’ve got right now.

And cost is obviously the thing that’s probably most. Cost and value are the two things that like the executives, the decision makers, your stakeholders are thinking about when you’re delivering a data project, the potential of the value. But what you need to do is get there in a way that makes it ROI positive and that’s keeping your costs down.

So you don’t want to be wasting some of that precious margin on using the wrong type of technology. 

Adam Sroka: I’m conscious of time. It’s been a really interesting conversation. I guess I’d always like to just ask a couple more things. If there are young budding analysts in the audience that want to follow your career path or work in a similar space, have you got any advice for things they should be looking at or doing?

Rosina Norton: Yeah, I think as the kind of data industry matures and especially how data is used within. Energy industry, which is quite often for forecasting or O and M, which are what I would call like production applications. I think for any aspiring data analyst or Data scientists getting some understanding of the foundational kind of software engineering practices is an absolute no brainer because I think one thing that for me differentiates a really good analyst or a really good data scientist from a very talented but maybe not quite up there an individual is someone who’s thinking about repeatability, defensibility, what’s my auditing here.

So that if you are running a piece of analysis, or you are building an ML model, you’ve got those engineering best practices in place so that you can have version control, you can say, Hey, those results, which were delivered six months ago. They were derived from this version of the data and this version of the data set and understanding why that’s important and how to implement it.

And I think that’s particularly important if you want to go into machine learning. We’re obviously seeing the kind of ML engineering and ML ops professions becoming more formalized now, which is absolutely fantastic because I remember putting my first ML model into production and it was me and some really talented software engineers.

And we almost like organically built. Like a feature store and experiment tracking and production, like inference monitoring, but it wasn’t, you always felt on a back foot because you knew that was missing, but you couldn’t quite formalize what that was or why. So now we’re seeing that come out in MLOps and ML engineering.

It’s fantastic. So I’d say to anyone looking at the data science career path, look that up. Even if you don’t want to be an ML engineer yourself, even if you just want to be the data scientist who’s creating. the models, just knowing the implications of that and actually how your product’s going to be used out there in the wild.

What are you actually forecasting for is I would say, indispensable. So yeah, if you’re coming at it from a purely analytic data science background, making sure that you get some exposure to software engineering best practices, and that might not even be for production, like models that could even be for your data warehouse.

So for example, I’m sure lots of people in the data industry have worked with dbt, which I absolutely love. I think it’s an absolute game changer for having really high quality data sets available for even just reporting via AI rather than data science. So I think, yeah, that would probably be one pointer.

And then the second one would be understand the business, don’t stop talking to stakeholders and understanding what are you actually trying to achieve? I think it can be really easy to get sucked into the weeds of what a technical aspect or some ambitious project you’re trying to implement. That If you don’t ground that in real kind of business context or what are you trying to achieve?

How do you measure the value of this? That could end up being useless if you deploy that report and no one uses it or you have that Machine learning algorithm, but you can’t defensively say why it’s better than the existing data practice you have in place. So I think, yeah, really grounding it in the current business context.

Adam Sroka: Absolute gold advice. Thank you very much. I’d say that probably one of the most impactful things I did early in my data science journey was read a couple of software engineering textbooks to try and figure out like, What that world was all about. And even if you’re not intending to be a software engineer, I’ll just echo what you said.

It makes you better at talking to the other parts of the value chain, like the other roles and understanding their concerns and why things are a certain way. Look, really appreciate your time today, Rosina. One last thing I always like to ask before I let anyone go. If anyone in our audience wants to find out more about you, or if you’ve got anything you’d like to plug or promote, where should we be sending them?

Rosina Norton: Yeah, I would say the best place to find me is on LinkedIn. I can post some general thoughts about data and clean tech. So I’m not trying to be an influencer or anything, but I think sometimes there’s interesting stuff I come across through my work. So on LinkedIn. 

Adam Sroka: Amazing. Look, thank you very much.

It’s been an absolute pleasure. And yeah, all the best. Hopefully get to catch up again soon. 

Rosina Norton: Yeah. Thank you so much, Adam. 

Outro: Thank you for tuning in to the Hypercube podcast. If you enjoyed today’s episode and want to dive deeper into the AI in the energy sector, be sure to subscribe so you never miss an episode.

We’re always eager to hear from our listeners. So if you have any questions, any topics you’d like us to cover or feedback, you can reach out to us on LinkedIn or check out our website at wearehypercube.com. And remember, the energy landscape is evolving fast. Stay ahead of the curve for the latest strategies and solutions right here on the hypercube podcast.

See you next time.