Hypercube Podcast Transcript

Title: Practical AI and Data Science to Deliver Sustainability

Host: Adam Sroka
Guest: Kevin Naik

Intro: Welcome to the Hypercube podcast, where we explore how companies in the energy and utility sector leverage data analytics to make smarter decisions and accelerate business growth. I’m Adam Sroka, founder of Hypercube, a strategic consultancy that supports asset owner-operators, traders, route-to-market providers, and energy services companies to unlock the power of data.

If you’re interested in hearing real-world examples of how data and AI are advancing the energy sector, this is the show for you. 

Adam Sroka: Welcome back to the Hypercube podcast. And today I’m joined by Kevin Naik. Kevin, would you like to introduce yourself and your background for the audience? 

Kevin Naik: My name is Kevin. I’ve done my PhD leveraging AI in energy sector, where I’ve worked on district heating and the thermal side. Of the energy, which is one of the energy guzzling units in UK.

In the process, I’ve worked with companies like Siemens MLM, Clean Energy, Equilibrium, and now working as an Head of Data Science at Intellium currently. 

Adam Sroka: May I ask what was your PhD title and did you love it or hate it as much as I hate mine?

Kevin Naik: So I enjoyed the journey because I still embrace that journey right now.

And I was also lucky that I did it in UK because I was working on two European projects. So I was way too lucky. In the sense that Simmons sponsored my PhD with Nottingham Trent University at the same point of time. Simmons had an European project and Nottingham Trent also had an European project which allowed me full of resources and try more research.

If I may say in that sense, I enjoyed my journey. I actually created quite a positive impact as to say, because It led to my career path very clearly in the process and my title was innovating and writing new algorithms for district heating using the cloud technologies to improve the heat generation plant.

Adam Sroka: Okay. Yeah. Okay. I get that. It’s an interesting problem space. And we’re seeing a lot of innovation across district heating, like not as much in the UK as some of the Nordics and things like that, but it’s a really cool problem space. And There’s some interesting stuff going on. 

Kevin Naik: Yeah. So actually the government of the UK has been quite active and the UK has got a more than 1500 district heating network with Nottingham being one of the largest district heating networks with Nottingham council.

It was Really fun to understand how they operate because they’ve got I can say plant rooms, which are Way too old to the plant rooms, which are fully digitised and like a brand new Stations, which was very fascinating to understand that how this all things in a network still can work to give a seamless outcome to us as an customer, which are commercial as well as residential in quite a point.

It is governed by renewable energy, which is also one of the points I should say here that to have 85 kilometres of a network and to serve it as a renewable side of the things is a big ask. 

Adam Sroka: I didn’t know there were as many as 1500 actually, that’s really interesting. I need to go back into my research on that stuff.

I think there’s a lot of potential with them, like the rise of demand side flexibility and co-location of storage systems, thermal batteries and things like that, it’s quite an interesting space to play. I’d like to see how it develops over the next decade. 

Kevin Naik: Yeah it’s very interesting. So I can give you a very simple example: we talk about Tesla walls. We talk about a lot of batteries. Lithium ion batteries and storage of the batteries, but apparently thermal storage as an energy is way to cheap than storing the energy as in battery So there is an innovation that a lot of people are doing that Can we use thermo thermal storage, which is one of the big energy guzzling in residential areas, in commercial areas in such a way that we get heat throughout the day with an optimised way of Doing things I saw one of the good research happening recently I think it’s a nottingham only one of the residential partners they have partnered with birmingham university and they are using air source heat pump and thermal storage in the house and PV panels, so the combination of those two and making net zero Homes, so which is very interesting thing similar to that.

There is one more residential scheme that I can touch upon which I worked very closely with was near Irvington where they actually created a passive home with only solar panels and the reading shows that actually they can go off grid at some point of time. So they have achieved the energy efficiency massively that they want. They say that they can go off grid, but it is still far from it in real life because sometimes at nighttime, you still need a grid to support electricity there unless you put power on it.

Massive batteries which is a cost effective solution. No. 

Adam Sroka: It could get, we’ll get right into the weeds of that. So we’re actually building for a customer a platform that does that home energy management optimization piece for thermal storage, heat pumps, and solar. So the engineer that’s almost in production, ready to go. And it does. Yeah. It’s interesting integrating different hardwares and things like that. So I think that’s going to be the way a lot of residential stuff goes. Especially on the, it makes total sense on the district side. Okay. Very interesting. Let’s step back a little bit. How do you think data and AI generally is going to impact these challenges or these domains, these use cases?

Kevin Naik: One thing is that a lot of people don’t realise that there are very mature softwares and simulators, which are available in the market. So they always go as AI being a very buzz word, as I always say right now, everyone wants to dive into it, but what systems, which we are talking about generally right now is a physical system.

So there are physical equations. They, the power readings, a lot of stuff. The question and the data and AI, which plays a very important role in the whole process is how to run them efficiently, how to run them so much in an optimised way. That’s where data and AI plays a very important role.

So let’s say for example, if I’ve got a residential building, I understand their demand from simulators. It can do up to only certain limits because it will just check the ideal scenarios and ideal way towards things. But once people start staying in that particular residential area, You get data from them.

Now you start understanding their behaviour and how they are using their energy. How you can adapt to that now there can be different cycles to this. So let’s say it’s a rented place where Rental is changing. So I rented it for one year and then I moved out of it And then you rented that same place for one more year now within three years You There are three different people using the same property.

Now this customer behaviour To understand them requires a short term model. It cannot work on very long term and historical models. If I go into very detail and also the landscape with a chat GPT or those kinds of things where people think, okay, every answer can be found from an open source or a general model.

This particular energy sector is one where that cannot happen. It changes so much. It is also stochastic in its own ways. So for example, someone can turn cattle on at night. What will you do then? That demand prediction is quite difficult for any model. And if we try to go towards something like open AI or chat GPT style of models, it is still not possible.

So I think the model in itself needs to learn a lot frequently from the data. And there needs to be a very fine line as well between that, because we can integrate as much as technology as we want, but we also need to find the good balance between technology, which is the resource and cost that we are spending behind that.

At the same time, what is the energy that goes in it? If I fire tons and tons of computing power, like Rolls Royce does for their aeroplanes for a residential building and burning so much carbon out there just to get my models, right? So I think that leverage and understanding it, that where we should stop and where we should see this accuracy works enough.

And for us, it’s still not explored a lot. People just think, oh, they don’t care. It’s just running cloud and, oh, you are getting my answers. I 

Adam Sroka: saw a statistic that I can’t remember, so I won’t quote it, but it was like the carbon emissions of Google, I think, have rocketed over the last year. The article was talking about how.

It’s probably to do with things like Gemini and BART, some of the stuff on GCP running these large language models and stuff, and I was really surprised to see, and then I saw another infographic and one of these, I haven’t checked the sources, but it was the volume of oil needed to run a thousand Gemini queries versus a thousand chat GPT queries.

And it was staggeringly different how the energy efficiency between the two. And I think they’re whether or not those things are even real, they were just like things that were shared in passing of half remembered, but I think they’re going to start to be questions that we as an industry, for us being the energy sector, have to really ask and get to the bottom of tech generally and everywhere we need to start saying okay yeah, you can create thousands of images or write a book in 30 seconds, but should you, and what’s the real cost of doing that stuff, creating all this content? No one’s going to. It’s a really interesting space to play. 

Kevin Naik: Yeah, it is going to be coming like the sustainable practices that are also gone into a lot of the people. So if you know that there was one era where plastic was used like anything, then they started using recyclable plastic and things like that.

So I think that trend is going to come for sure. Yeah. Where people are quite aware about their sustainability practice, but still from a technology front, they are unaware of it. And once people start researching and once this infographic that you talked about would get a little bit wider, whether there is a truth to it or not, we still need to look into the source for that.

Yes. But that is a good awareness tool. I feel because of me being a technocrat, I still want people to run models as low as possible if they can train it once a month. I would like them to train every six hours. As simple as that, a lot of companies, unfortunately, try to do that because they have got customers to pitch into and they don’t go into the performance matrices a lot.

Big companies do, but I’ve not seen a lot of startups or a lot of SMEs still going into that particular thing. And they are just worried about the solution that they need. Today’s world is such that they need a quick solution. They cannot wait even For 30 seconds to get the solution. So i’m sure you might be facing a lot of clients when you build the platform that the latency should be Less than one second 

Adam Sroka: Again, it was a video by Andrew Ng about agents.

And he said, one of the things that we might have to accept is that we aren’t going to  get 500 millisecond response times from every application and actually give some of these tools time to generate an answer and refine it overnight. Or whatever might be essential to get the best results.

Whereas we’re als, and we talk to customers about it all the time, but everyone now is anchored to really high production quality web applications that are really responsive. They look beautiful. They’re really reactive. And that isn’t actually a hard requirement for a lot of use cases. It’s kind of nice to have.

It’s like the polish at the end. But every end user is like, oh, this is a bit slow or it feels clunky. And it’s a real requirement. 

Kevin Naik: Yeah. They always say, oh, this application is refreshing but they don’t understand that it is real time data that is coming in that has been processed and then thrown out to you as an answer which goes through processes. So it’s like they just feel they stare in front of the screen here. One bottle has been filled.

I should just know on the screen that the bottle has been filled in a real time situation. And there are a lot of non tech users as well. I feel those users Are associated with certain tech but they’re still not associated with tech like software and especially data science where I feel They just see that magic happens and they want more and more magic to happen and more quickly. That magic happens better.

It’s a general Association that I’ve seen with most of the clients that I have worked with. And that’s why I think this web app is also coming quite a bit for each use cases, people have now started building separate containers and just building the web app for that particular thing.

So it performs faster, giving that magic to the client. And then just keep on doing other stuff in the backend side. 

Adam Sroka: No, I don’t see that reflected a lot, actually, as well. And that makes the technical audience having trouble crossing the chasm a little bit, like changing gear a bit.

How do you make the magic happen? What kind of tools and technologies are you a fan of and do you use? day to day workflows. 

Kevin Naik: In general, a very open source tech guy, if I put it in that sense, then a closed tech, the, my simple reason for open source is that it gives you a lot of configurability possibility.

And whereas closed tech puts you in such a space where you feel comfortable, but then after some point of time, you also feel restricted. Whereas open source you can build your own wrappers and your things and can go ahead With those kinds of stuff. So I’m more of an open source tool user and also if we Like I can give you a simple example from my past experience is Tabio is a very classic visualisation tool.

It’s a fantastic tool. So they have built a very good tool. I cannot say that they have not built a very good tool, but it is only good for certain users. It is not good for each and every user. Second thing is that I don’t have enough flexibility which Me as a company or me trying to give something to the client with my Branding and my logo on it.

So I use the Flask maze web app. So it’s an open source framework that you can easily create Web apps and I try to promote that even internally in the teams as well, that if you standardise those codes, you can then basically utilise those kinds of things. There was a very interesting tool, I don’t know now if in industry that is used anymore or not, is node red, which was created by, and was last owned by IBM.

But I know that it still exists. It’s not that it doesn’t exist, but it was based on. Concept of Google, which was blocky. Which is you fix different blocks to gather and it was not exactly no code platform. It was flexible enough for people to write their own code and put it in the blocks and then connect the whole workflow and get your outcome.

There are now a lot of other tools like Apache Airflow, MLflow. There are kubernetes and there are a lot of open technologies that I still prefer. Open source Repositories are maintained by different groups of people. So at certain point of time they may decide Not to support that or they may just leave it at last possible Repository and then we find that Repository is not available At all when they try to reproduce that code that happens very often especially in data science application when i’ve seen junior data scientists taking up some libraries which is a wrapper of some standard library, but that wrapper has never been updated in last five years.

And then when we started decoding that code, it took us about three, four weeks to decode that code. And then we were like, Oh shit, this is just an horrendous mistake being made. So I would say even in open source use more standard libraries also look at their histories. Who are the contributors?

Is it maintained regularly or not? Because a lot of people just use it because it’s open source. Just use it. Also see the commercial licences. That would be my message as well with 

Adam Sroka: That’s a really good tech overview via some of the tools and challenges of designing these and building these systems I’d ask do you have any kind of top level advice for what are the key success factors?

From a client perspective, when you’re designing these solutions, any advice you can give? 

Kevin Naik: One key advice to start with is to understand that problem very clearly. I think most people have. One of the backgrounds. So they have either engineering as a background or computer sciences as a background there.

And a third category that I’ve kept on seeing now in people who have grown into a data science career in a sense is someone who has switched their careers. By learning data sciences and trade. So skills like upscaling, like the UK has got a lot of programs where they’re upscaling the data skills in general.

There are three categories of people. I think the most successful people are the collaborators who understand energy very clearly. At the same point of time understands the tech stack and technology which will go to solve that problem Very well, it’s difficult to find people like that very honestly So my key advice to everyone would be get a clear scope from client if it’s your client If it’s your internal team as well define your scopes Very clearly break it down into a small machine learning problem as possible and then try to tackle it in a most efficient way that you can tackle.

And just don’t go behind madly writing 30, 40, 000 lines of code. Just for the sake of writing and then saying, Oh, this is not doable. I think it needs to be done at the right point of time and in the initial stages. You may delay your process. I know this is something which every business wants a quick turnaround, but if you do your scopes very clearly, your turnaround time and your optimised time would be much higher than.

Expectations, I can rest assured. From two of the projects that I’ve done before. So yeah, that will be my key 

Adam Sroka: advice. Great advice. And yeah, a few points I definitely live and breathe by. I’ve got a whole company on the fact that you, it’s hard to find people that know energy and the tech.

So yeah, that was the dream for Hypercube but yeah, breaking things down into what’s the smallest possible bet that you can make. And how do I prove something and then move forward a step? I think that’s a mindset you really need to, Get into space. I wanted to touch on you were recently up here in Edinburgh and you were speaking at Euro Energy 2024.

Is that right? What were you talking about there and what was the response reaction like? 

Kevin Naik: It was an interesting event, actually, as a keynote speaker, what I touched upon there was how AI can be used to leverage sustainability. Energy system. So in that particular talk, what I talked about was few of the examples that I’ve already covered before, but I’ll just quickly cover one is district heating, how district heating can be optimised, then how you can go to net zero by using solar panels and micro turbines.

So an urban turbine and solar can get you to the net zero. journey out there and then how to make your production lines, your machines more and more efficient to deliver what it has been actually meant for and achieve that operational excellence can give you more production and hence can be more sustainable than how it has been run.

And lastly, the Energy centres where you have heat pumps. You have got boilers. You’ve got batteries, how to optimise those things using ai and get a good result. So it was more talking about some of the Results that you can see from my previous work, which was at equilibrium. So they’ve already published a few of the case studies where we were able to reduce Energy 10 to 12% for a lot of our clients and increase the operational excellency by 10 to 15% just using data and machine learning there.

And then the other side was the results that we got from the ment project, which was an European funded project, which actually I started my PhD with, which was in nothing And now, which is actually good news. About a month ago in a newsletter was that the same project where we did the pilot for 27 homes has got another grant to go.

And do it for 300 homes, which was the initial plan to do. And all the learnings that we got from these 27 homes, positive and negative, would be implemented in the next phase that they are going ahead. So yeah, that’s what I presented at Euro energy 2024. It was good to be called as a keynote speaker to establish a few researchers that are more established than we were.

Adam Sroka: No, very cool. Wow. Yeah, I’ll have to check out the talk and maybe see you there next year So just one thing I’d like to ask everyone before we wrap up if People want to find out more about you or if there’s anything you want to promote or plug would you like to share with the audience 

Kevin Naik: To find me, Linkedin is the best contact.

I always accept everyone so everyone is more than welcome to add me on LinkedIn, whosoever listens to this podcast. And any questions starting from how does work in academics, to how to work in the industry, how to grow your career from being a data scientist to head of data science.

Apart from that, my interest is in robotics, artificial intelligence, and IOT. So any talk that anyone wants to do over a coffee or anything, then more than open to do that. 

Adam Sroka: We’ll be sure to share that. Look, it’s been great to speak to you. Kevin. Thank you once again And yeah, all the best at Intellium. 

Kevin Naik: Thank you, Adam. It was nice speaking to you. I’ve learned a lot about your company during searching and coming for this podcast. I saw a lot of amazing work that Hypercube does at any point if you need any support from Intellium or from me.

Please feel free to connect and more than happy to hear from you. Few more challenges and things offline as well. 

Adam Sroka: Thanks. 

Kevin Naik: Thank you, Adam. 

Outro: And that’s it for this episode of the Hypercube podcast. Thanks for tuning in today.

If you have any questions about the topics we covered, you can reach out to us on LinkedIn or check out our website at  wearehypercube.com. You can also join Beyond Energy, our Slack community of data leaders from the sector. There’s a link to sign up in the episode description. We’re just getting this show off the ground, so if you like today’s episode, please leave us a rating, review, or subscribe wherever you get your podcasts.

It all really helps. See you next time.