Jan Drgona on Solving Problems with Energy Sustainability in Buildings Using Scientific Machine Learning and Engineering

Transcript:

We can build buildings with different materials. We can build with concrete, we can build with bricks, we can build with wood. They will have different thermal properties. Most of you are talking about thermal mass and insulation or thermal resistance of the material. They really play a significant role in fighting this second law of thermodynamics. Hi. Welcome to the Science Fair podcast. I'm your host, Susan Keetley. I'm a PhD chemist, writer, and I love talking to scientists. On the Science Fair podcast, I aim to bring you conversations with scientists doing fascinating, cutting-edge work on all kinds of interesting phenomena, ranging from physics to chemistry to biology, and even the nature of science itself. Tune in every Monday for a new episode. For each scientist we interview, first we'll release a mini-episode that connects what the scientist is doing with what's happening in the high school science classroom and then the following week, the full-length interview. So come along. And tune in for some science fair. Our guest today is Jan Dragoña, who joins us from Johns Hopkins University. Jan is an associate professor in the Department of Civil and Systems Engineering, and is also at the Ralph S. O'Connor Sustainable Energy Institute. Jan's research focuses on energy management in buildings, and he's working on developing scientific machine learning methods to model energy management, which turns out is very complicated. And our conversation today is going to touch on a lot of pieces in the problem of making energy use in building sustainable. Namely, the complexity inherent in the buildings energy use, what scientific machine learning is and how it can help, and problem solving with an engineering approach and mindset. So Jan, welcome to the show. [MUSIC PLAYING] Thank you very much, Susanne, for having me. So I'd love to start at the beginning. Can you tell us how you became interested in science and decided to pursue a career in it? And how did you decide to specialize in machine learning for modeling complex energy systems? So my interest in science and engineering in general became obvious in very early age. The specialization came way later. So ever since as a kid, I was drawn into NC Club Media books and engineering manuscripts and everything I could, playing with Lego bricks and building systems out of that. I was really fascinated how we can build something. Out of small components. And the moment when I realized I want to become a scientist as a profession was probably during the high school, during the chemistry classes. We had wonderful chemistry teacher who was a very hands-on, who was always showing us these cool experiments in the chemical lab and things that have been exploding, changing colors, generating very hideous smells. And I was really fascinated by these transformations that are happening in science and in nature around us. And I wanted to know how these things work. And in parallel to that, I also discovered that I have a talent to understanding these transitions in abstract mathematical language. So I was combining the mathematics and chemistry into my passion while also spending significant amount of time as a teenager playing computer games. So I figured, I probably should do something that combines computers, math and chemistry. So I decided to do my bachelor's in chemical engineering, but with a special focus on industrial automation that combines programming of cyber-physical systems, where you have the cyber part, the digital component, and you have a physical system. So this was very fascinating part for me that I didn't have a grand strategy in mind where I'm going to end up. I was exploring different branches of engineering during my studies. After my bachelor's, I did master's thesis at the electric engineering department. They'd my PhD in control theory, focusing more on computational and mathematical aspects, and transitioning afterwards as a postdoc to Cal Lumernick Belgium, focusing on mechanical engineering aspects. So it was very winding path that was driven mostly by very lucky encounters. Can you tell us about one of those encounters? One of the encounters was through my PhD with my academic advisor, where I knew that I wanted a PhD, but I didn't know in what topic. So he was the person who pointed out to me to the building control as very complex and interesting problem in the control theory. And that really hooked my fascination because it required multi-disciplinary approach, not only knowing the math, but also understanding the physical system. So that resonated with my desire to work on some practical systems, not only some academic examples. And as a part of this exploration, I was attending one of the scientific workshops on a topic where I had a chat with another PhD student from other universities during the coffee break. And we have been conversing about the problems on the energy efficiency of buildings, and he shared that he is working on the high-fidelity modeling of physical processes in buildings, and he's looking for someone who has expertise in optimal control methods that he can apply. And I was looking exactly for someone who has deep expertise in modeling because I was specializing on these controlled theoretical approaches. So it was really a certain deepest encounter that really defined my career and really changed my life, because after this encounter and writing papers together, I joined their group as a postdoc, and wonderful three years in beautiful city of Lumin in Belgium. During that time again, we had collaboration with national laboratory in the United States, and specifically at Paswig North, as national lab in the beautiful Washington. And through this collaboration, I again transitioned from Europe to US. They had a national app for another five years at the physics and computational sciences division, where we have been building real software solutions for these energy optimization problems in buildings and beyond, that's really materialized from early studies and prototypes into multi-project portfolio that has been supported by the department of energy with overall $20 million as support. So it was very satisfying to see how the passion from the high school was transitioning to some real-world tangible impact. Yeah, and I love the story of meeting someone. It was at a conference. I just was speaking to a scientist yesterday who has had a 20-year collaboration that stemmed from something similar. He was giving talks at a conference almost 20 years ago, and someone said, "Oh, I really love those talks." And then the head of the conference connected the two, and they've published over 100 papers together. But it just, there's something so special about the in-person interactions that you might not expect that fosters this. Absolutely. We are still people, and even as a science, we can seem as a hard science is working on engineering problems, but there are people behind those solutions. Yes. And it matters a lot whether we like the people who are working with, whether we are aligned on this solution, but also as a human beings. Yeah. That's very, very important aspect. And now you're at Johns Hopkins. How long have you been at J.J.? It's my second semester. I started January 2025. Wow. Wonderful transition and part of the RALPHO Connorsystem by Energy Institute. That is world-leading Institute and Sustainable Energy. It's an amazing place for interdisciplinary collaboration, basically every single department from Whiting School of Engineering has their representatives. And we have wide range of applications that we are focusing on. It's not only energy system of buildings. We are interested in optimizing complex power systems, designing new materials for more efficient batteries. We have faculty who are working on carbon capture from fundamental chemistry to startups. They are actually being deployed in the world. So it's a really, again, unique place to work from basic science to applied science to actual translational projects in the world. That is really exciting. Well, I'd like to talk about the problem you're working on, which is buildings and their energy use, something like 40% of global energy use is from buildings. And in the United States, 70% of electricity is used by residential and commercial buildings. But the systems used to control the energy are inefficient. So can you tell us more about that? First, why do buildings use so much electricity? And second, why are their inefficiencies and what's the problem-solving opportunity here? Yeah, absolutely. Buildings are very complex systems. They have hundreds of different subsystems. So it's not system. It's a system of systems. Now, we are talking about gas furnaces. We are talking about heat pumps. We are talking about pumps that circulate the air or water throughout the building to facilitate the heat transfer from our medium to the indoor environment. And the real purpose is to keep us in a comfortable environment during whatever weather conditions are out there, whether it is freezing cold or very cold. is baking hot, we want to have as human beings some comfortable stable environment around 74 degrees Fahrenheit roughly, but it's not only temperature, it's also humidity and indoor air quality that really matters for our productivity and for our health reasons. So why it is expensive? Because these conditions are not stable in the environment. We have weather changes, there are also temperature changes in day and night, so we need to condition buildings for this table indoor air quality and temperature. That's costly because it requires running the heating ventilation air conditioning systems, essentially 24/7 in a way how it is being done these days. And not to going to too much technical details, the complexity is really when multiple buildings or large buildings are coming to place because when you have single residential home, you have what is called a thermostatic controller that measures the indoor temperature and has very simple rules that decides whether it should heat or cool or do nothing. So when we are talking about single apartment building or single residential house, that's sufficient. But now when we are talking about medium to large scale office buildings, we are talking about hospitals and now data centers, they are massive users of electricity at the end and they have stringent requirements on the indoor temperature that in practice we don't optimize yet because the current infrastructure is very outdated. It has been built in 70s, 80s based on the boiler-prays from process control industries where you could have operators, engineers overseeing the operation of the chemical plant 24/7 and doing small adjustments to optimize it manually. We cannot do that for buildings, right? Because that would be too expensive. So a lot of techniques and theories have been developed how to do it automatically in an optimal way and we know how to do it in theory. But to do it in practice, it's very expensive because it requires modern IT infrastructure, it requires relatively performant computers and sophisticated algorithms that require advanced engineering degrees. So these are the bottlenecks that allow us to optimize the operation of the heating ventilation and the conditioning systems of buildings. So instead of that, we are using outdated decision-making rules that are based on so-called rule-based control systems where typically HVAC or building management system engineer who has lifetime experience in designing these rules to make those rules manually for every building specifically. And you can imagine single human cannot really optimize system with hundreds or thousands of degrees of freedom. So to give you an example, when we are driving a car, we are doing decisions at real time and these decisions are typically in orders of 12 degrees of freedom. We are going left or right, yeah, we are speeding up or slowing down, we are changing the gears and now even automatic transmission, like you don't need to worry about that. But still driving a car is a complex decision-making task. They require our full attention. You cannot drink or you cannot check your phone and we need to go through the driving school. For 12 degrees of freedom, yeah. For 12 degrees of freedom. And now we have a building that has hundreds to thousands of degrees of freedom. And we have a person who tries to synthesize these decision-making rules in few lines of code and leave it as this. So you can imagine that these static decisions are not adapted for dynamically changing environment. And I would imagine there's just a lot of waste inherent in that because this very simplified model can't get at all of these nuances of each room or each collection of rooms and their needs. Exactly. And because of the lack of sophistication in our decision-making capability that can be actually practically deployed in buildings, those rules are designed to be very simple and they are designed to be conservative. But not conservative in terms of energy conservation, but in terms of delivering stable indoor environment. So basically they are running all the time even if they don't have to. So for instance, we know that let's say Monday there is a state holiday, no one will be in office, but our Rube Assistant doesn't know that. He will be still running all the time. The Rube Assistant doesn't know that tomorrow will be holiday. So I don't need to heed the building because what my rules are telling me, but that's what is happening. And that's why very often during the summer, when you go to large conference room, it's freezing. People are shivering and they are bringing their. You have seen it probably yourself. People are bringing their sweaters or even wars running electric heaters during summer in the office building because it's freaking cold, right? That's amazing. I think if I worked in a cold office, I would have a space heater. That's happening. Literally our admins here, they have the space heaters and next to them running frequently. So these are the real sources of energy inefficiencies that are stemming from the complexity of the decision making in building a conference space. And before we get to how you're attempting to solve this problem with scientific machine learning, I kind of want to couch what we've been talking about in terms of the second law of thermodynamics because I like to connect to what we talk about on the podcast with very concrete things, kids are learning in school. So, you know, I'm sure you see examples of the second law all the time. You have hot or warm air going to a cooler place. The fact that heat tends to flow in this direction, how does that make what you're doing more difficult or maybe in some cases easier? How do you kind of incorporate the second law of thermodynamics into thinking about this problem? The problem is that we are fighting inter here, right? We are fighting the second law of thermodynamics that is telling us that the heat will flow from warmer body to the colder body and there is nothing we can do about it. So, if there is a cold temperature outdoors, you are doing the winter and you want to keep nice cozy warm indoor environment, there's nothing we can do about our building wanting to get cold. There will be always energy dissipation in the environment. There will be always a heat losses. So, that's why it's also costly from the heating ventilation air conditioning perspective because the wider the gap, the more energy we need to provide to the system to keep the indoor temperature out of the equilibria with outside environment. So, we are forcefully keeping our building indoor environment to be where it wants to be equal to the outdoor environment. But a lot of depends on the building properties. It depends how buildings are built and it depends how buildings are operated. So, I touch base a little bit on how buildings are operated. So, how about how buildings are built? They will have different thermal properties and most of you are talking about thermal mass and insulation or thermal resistance of the material. And they really play significant role in fighting the second law of thermal dynamics. So, we can use these somehow into our advantage. The first thermal mass is how much energy can the material store. So, the buildings that are built from bricks will have more thermal mass than wooden buildings. And we can really use this to our advantage because they can absorb heat. They can store the energy. So, let's say during the day there is sun shining on the wall so it hits the bricks or we are heating the indoor temperature in room from inside with temper the bricks. So, they can dissipate this heat during the night. So, what they're effectively doing, they are smoothing out the temperature fluctuations. Technically, they are called so-called low-pass fiertel filters. They are filtering out the high frequencies and they are passing the lower frequencies. And this is very important because sometimes it's not only the overall energy demand we care about, they're also the peak demands that we care about. And they can be get very costly especially in today's cover grid. So, that's something that the thermal mass can help us with. The second aspect of the building material is the thermal resistance. That really resists the heat escaping the indoor environment then outside the environment. And we can use these to use less energy deliver to the building to provide the desired overall temperature. So, combining the thermal mass and the proper insulation of the building, so higher thermal resistance, we can operate these buildings more efficiently even without active intervention. Just by avoiding the heat to escape and avoiding the heat to have high fluctuations. And the third example, what I want to share is the type of heating ventilation air conditioning systems we use. Play the significant role. Why electrification is very promising and potentially revolutionary technology from the perspective of energy efficiency? Because of what we call coefficient of performance, which measures how efficiently we are heating or cooling a system, in this case a building, using our HVAC component. And the simplest way is to use energy conversion approach like burning fuel using gas furnace or burning wood or coal in the fire stove that will have at most for gas furnaces. 90% maybe 95% of coefficient of performance, which means for one energy in the fuel, we will get 0.9 energy units in delivered heat. So it's coefficient of performance below one. Yeah. And this is because we are burning fuel. We are releasing the heat from breaking the chemical bonds in our fuel. So is a chemical transition process. But in terms of technology that is called heat pumps, we can get to the coefficient of performance of 3 to 5. Oh wow. How is this possible, right? Well, it's possible because we are not burning anything. We are not releasing the energy based on breaking chemical bonds. We are using the electricity to operate the heat pumps to effectively transfer the heat from the environment to the indoor conditions. And what is a heat pump? Well, everyone has a heat pump at home. It's called a fridge. It's essentially a heat pump, right? It keeps the inside of the fridge cold while dissipating the heat to our room. But it doesn't do it by burning anything. It's just doing it by compression and evaporation cycle that really smartly utilizes a law of thermodynamics to bring the heat from one outdoor environment, release the heat in another environment, and then cool down the medium to heat up in the other environment again. So it does that very smartly moving the heat from one space to another space. And that allows to operate with 3 to 5 coefficient of performance. So for one unit of energy, you will get 3 to 5 units of heat. So it's a marvelous technology, very, very environmentally friendly that has the potential to provide as much energy savings if deployed at scale. That's amazing. I had no idea that the coefficient could be so high for a heat pump. Yes, yeah, if operated efficiently. But, uh, right. And that's sort of where your work comes in. But before we go there, I'm just going to read something for high school students who have schools that follow something called the next generation science standards. So I wanted to let listeners know that what we just talked about matches with standard PS3-4, which says that students should be able to plan and conduct an investigation to provide evidence that when two components of different temperature are combined within a closed system, the transfer of thermal energy results in a more uniform energy distribution among the components in the systems. So it's a little verbose, but I think if students are tasked with writing about how you would plan and conduct such an investigation, it seems like buildings could be a great setting for how you would look at that. This spiritual, one of the solutions that can really help in energy savings, again, using now the second law of thermodynamics to our advantage is, for instance, if we know that one part of the building is warmer during the day because of the solar irradiation gain, the sunny shining to the south side, let's say, and then we have the north side that is a little bit cooler, but we have the hydronic system that is interconnecting these two. We can recirculate the water to transfer the heat from the warmer to the cooler end of the building without actually spending anything on burning the fuel. So again, we can utilize the laws of thermodynamics into our advantage to operate the buildings way more efficiently, but it comes with increased complexity to actually being able to represent these physical laws in mathematical equations, so we can make optimal decision by simulating these mathematical equations in so-called digital trends. Part of what you're using to address this problem is something called scientific machine learning. I am familiar with plain machine learning. Sort of my crude way of thinking about it is you train a machine to look at a lot of information and recognize patterns, make decisions, it might be able to make a prediction. Can you talk to us about scientific machine learning? Why is it different and how can it be helpful here? Let me first refresh the machine learning paradigm, which is a method to make machines, to make predictions without explicitly being programmed. Well, traditional programming is so-called imperative. We take inputs, we write the computer program with exact instructions what it needs to do with those inputs and it returns the outputs. Machine learning does the opposite. We provide the answers we want to get and we figure it out how to transform the inputs to get those answers and that computer model has different forms of what they call machine learning architectures, whether it is like linear models or now very popular neural network models or artificial neural networks, they are not biological neural networks, but these are basically pattern matching mathematical models that allows us to tweak a lot of parameters to transform these inputs to outputs and they are marvelous technology because they are driving a lot of changes in society these days when everyone is using chat GPT. This is machine learning model that has been trained essentially on the whole internet to match the inputs our prompts to the outputs, their responses and it's fabulous how it is working in such accuracy based on purely pattern matching technologies. But when it falls short, is when we want to deal with physical systems that have hard constraints, that have physical laws that needs to be satisfied either from performance perspective or even safety perspective. Their systems are so-called safety critical. If we violate certain conditions, these can go wrong, very easily like chemical reactors or autonomous vehicles. In buildings, we also have some constraints which we need to satisfy because our equipment runs at a certain limits. These are not limits that we can just ignore. So here where we need fundamental understanding of physics to enforce those limits. And this has been decades of research in engineering domains like mechanical engineering, electrical engineering that understood the fundamental physical laws and utilized them to build engineering systems like heat pumps or understand how to design buildings with favorable thermal properties. So one of these examples was my project with my former colleagues in Belgium where we had this high performance modern office building that was equipped with so-called thermally activated building structures. You can think of it as a floor heating basically embedded in walls. So each wall, concrete wall had pipes of hydronic loops that was water recirculating. So when we say thermally activated, we can decide whether we cool or heat not only the floor or the indoor air but the whole building structure. And on the top of that, it had very efficient heat pump coupled with borehole which is basically deep underground loop of an underhydronic loop that allows us to exchange the heat not from air but with the ground which after certain threshold has a stable temperature across the air. So we can get to very very high coefficient of performance with these buildings because we can operate on very small temperature differences where that's where these heat pumps operate the best. Was this a real building or this was a model? This was a real building. We conducted the real field experiment for several months running high-fidelity building models and advanced control techniques over several months was in 2018. And back then we showed if we know the physics, we can build high-fidelity digital twins. We can use advanced math and optimization theory to build these really smart predictive decision-making policies to take into account not only the current state of the building but how the building is being used by the patterns of the occupancy by taking into account weather forecast, taking into account really detailed heat transfer of building simulations to achieve 50% of energy savings while simultaneously improving the indoor air quality for people inside. And this was a real building, real case study. It's incredible. But there is a book. This project took one year to design and deploy by six people with advanced engineering degrees. This was pure engineering. There was no machine learning in the book besides predictive models. And here where scientific machine learning comes into play. So what is scientific machine learning is a new methodologies that systematically combine machine learning capabilities. So it ability to match patterns to data and physics equations and high-fidelity simulations to represent the reality we live in, which is a physical reality. And when we combine these two paradigms into one algorithmic and software framework, it really enables us to learn from smaller amount of data because we are already enforcing certain physical laws. Like we know energy conservation law. We know the heat transfer. We don't need to discover it from data. We can say that these models needs to respect that. I think I'm understanding this now. You would want to give machine learning patterns to learn from when they might be unpredictable or they would do things that otherwise you wouldn't know. But in this case, we have scientific laws, we have equations, we have things that we know already. We don't need to train an algorithm on a library of them, like you can just feed it in. Exactly. Yeah. That's what we are doing in my group. How do we systematically embed the physics and engineering insights into this machine learning machines that can now provide us with more adaptive behavior as opposed to pure engineering approach that it's somewhat rigid in terms of designing the systems because everything needs to be designed by hand by knowing like engineering principles. So it really combines the best of the two worlds, the guarantees of the physics and the adaptability of machine learning. So now we can have high fidelity simulation models and these advanced smart control techniques that don't have to take one year to design and deploy. Then maybe take a one month or even a couple of weeks. So that's what we are really pushing by developing new methods and software tools to democratize the technology that is not restricted to handful of PhD graduates. But it can be deployed at scale with people that don't necessarily need to have advanced engineering degree in machine learning and control theory and mechanical engineering physics. It sounds so exciting. Yes. So we have very promising future ahead of us in this space because the so-called scientific machine learning is not only revolutionizing building energy, but is having tremendous impact across system energy applications from power grid operations, from battery design and beyond that in advanced manufacturing, in material science, in the in forma. So it's becoming very trendy topic on the intersection of AI and traditional engineering domains where we combine what we already know and we learn already only things we don't know or things that change. Right. The things that change. Yeah, that's so interesting because in the buildings you have a whole set of things that they're following laws of matter and nature and then you have, you know, like you said, the occupancy, what might the weather be, that sort of thing. I'm just so curious, like what do you think would be the most complicated building on the earth? What would make a building complicated? Right? Is it like height or it's more of its like spatial reach? It's a great question and it comes up. A lot of things make systems complicated. So the size is absolutely one thing because now we have more energy you need to deliver to the system, but also the shape of the building, the topology, how individual floors and the rooms are connected because that defines the heat transfer, the type of the heating, ventilation, air conditioning systems, and also the requirements on the indoor quality. So I would argue that probably the largest buildings that needs to have multiple cheerleaders plans to cool them during the day are the grand challenge or now the data centers that are being built at a rapid pace, having massive energy demands for cooling of very heat intensive energy intensive compute that needs to operate relatively stable temperatures not to damage the computational hardware. So these are next frontiers in our domain, how to be effectively handled these massive loads efficiently and how do we also provide services for the grid. So we are not only thinking about single building as independent system because I mentioned that building itself is system of systems, but when we think about grid level, the whole grid is system of systems. On the load level we have multiple buildings, we have data centers, we have residential buildings, we have commercial buildings, we have hospitals, they all have different requirements, different time of use, different peaks, and when it they all aggregate together, they need to be served by optimizing our dispatch decisions on the grid level for our generators. So the grid is in the stable load and we optimize the transition of the and distribution of these electricity resources in the optimal way. So again, here the combination of physics and machine learning can play a tremendous role. Can you share some recent successes? And also, I know it's so hard to do this, but I wouldn't say where will we be in five or ten years, but what would your hopes be in terms of maybe how just everyday people this might start to affect their experience or the buildings they're walking into? My hope is to deploy these advanced control strategies within ten years, at least at most data centers because they are most available to change because they're built right now. We can influence how we build them in terms of the design, in terms of how we choose the IT infrastructure and how we deploy these advanced control strategies. So every time when there is something new, there is opportunity for disruptive change. There is my hope that data centers are kind of disruptive for our grid because they are bringing a lot of load demand, but at the same time, they provide opportunity for modernizing how do we operate complex buildings because essentially there are complex buildings. Because one of the bottlenecks in a current building domain is it's very conservative. Buildings are built to last. You build a building and it will be here for decades, not for centuries. It's a good thing because we want to build things that last, but it means that the change takes way longer from the regulatory perspective, but also from the historical perspective of the infrastructure. The current infrastructure vendor companies, they don't have a lot of incentives to change the status quo to innovate here because, frankly, from the monetary perspective, from a financial perspective, still heating and cooling of buildings is not as expensive. What is more expensive are people. So that was also one of the anecdotes that I can share when we have been doing these experiments in a building with Belgium. It has been commercial buildings around 100 people coming to the building every day. And first days, we have been experimenting with things. So we had a couple of days when we essentially screw up, right? Like we delivered too much heat and it was too warm or it was too cold. And we go to call from the building owner and say, "Look guys, if you keep doing this, we are shutting you down because I don't care about energy efficiency here. I care about my people not complaining." And this is the truth, right? The literature. So we need to provide energy efficiency while having the primary objective in mind, which is to either keep the compute hardware at the stable temperature or keep people not complaining. But the good news here is people keep complaining because the way how we control these buildings right now is with these simple rules, which imagine that you are driving a car, but instead of looking ahead, you are looking in your rear window and you're having one eye closed. So basically, that's how we are operating buildings right now. So what we are trying to do is to actually start looking ahead. So we can predict what is happening in our building, utilizing modern sensors. So we know the occupancy patterns. We can utilize the more accurate weather forecast to act proactively. So we have any know that there is a cold front coming and it will be colder. I can preheat the building to avoid big peaks of electricity that can be expensive and also disruptive for the grid purposes. I can utilize all of these second law of thermodynamics tricks that we mentioned before. And in synchronicity, we can really achieve high energy savings up to 50% of the dimension. So that would be my hope that we can demonstrate this not only in academic papers and the case studies that last for a couple of months, but we have more and more buildings that actually operate like this 24/7 for the case to come. Very exciting. I also, I'm thinking about the people complaining. I feel like there's a generational something at work here because when I was growing up, you know, when it was winter, you had to just put on more clothing and summer was really difficult. It was so hot all the time and my daughter went to a sleep away camp two summers ago where it was an air conditioned. And I thought, oh my gosh, I think this is the first time in her life that she hasn't been like between 68 and 72 degrees. Like, you know, it's like we have like so little patience for discomfort. I mean, yeah, it's excellent, excellent comment. And yeah, do you think about that too? Like, I grew up in Europe. There is no air conditioning mostly everywhere, right? So yeah, it's a very different perspective on expectations on comfort in general, like not only in our equality. Like, once you have certain standard of living, certain standard of comfort, it's very hard to experience that discomfort. And I firmly believe even though we have all of these luxuries and comfort in our lives, it's good for individual is good for society if we go out of our comfort zone and get used to it and brace that. Go to nature camp in the rain, camp in the freezing temperature to experience how life can be. Right. And appreciate what we have and appreciate that what we have is not even, it's not, that doesn't come out of nothing. It's cost it, cost money, it costs energy. And these are complex engineering systems that require a lot of care and maintenance. Absolutely agree. I think although I'll think about big complicated buildings differently after this conversation, just really appreciating everything that's going on behind the scenes to make it somewhat comfortable. Absolutely. And yeah, buildings are not the sexy applications that that you will have on the front page of science or nature where you have these high-dynamic drones or autonomous vehicles or games that capture our imagination and interaction. Buildings are kind of boring and slow and we are always there. So we kind of learn how to ignore them. - I have two questions for you. So the first is just reflecting on having a career in science. What do you enjoy most about working in science? - Probably the people that I'm going back to how important the relationships and social aspects are. I'm not only in science in general life, right? We are social human beings and we want to feel belonging and appreciation and the scientific community is very unique. It's very multicultural, it's international. It really changed my life profoundly. I grew up in small country in the middle of Europe in Slovakia in a very small town. So when I was growing up, my worldview was somewhat limited. And when I was start traveling through student exchanges and more and more traveling during conferences and I spend living in different countries across Europe and now different states across US, you meet different people with different backgrounds, different value systems, different cultural habits. But what you realize we are all humans. We all strive for certain fundamental things and we can learn how to get along. And I think the scientific community is really a beacon of hope for humanity because we learn how to do that. When you go in the US, academia, or also like many other countries, you really see this interdisciplinary and multicultural spirit of collaboration and appreciating different opinions. So that's one thing I would definitely highlight. - That's such a great perspective and it makes me reflect. I mean like you, I grew up in a small town. I grew up in a small town in Massachusetts. And you know, from the time I was in college, I was doing summer internships in labs. Then I decided to be a science major. And I think I took it for granted. How, I mean the first summer I was in a lab, I think there were five languages spoken in the lab. I think I took it for granted, like how much that exposure just came to feel normal to me. To always have people, you're hearing another language, you're seeing like the interesting things, people are bringing for lunch that doesn't look like your lunch. Everyone has like an interesting story. That is something really fantastic about science. - And that's very enriching from the human perspective, but also from the scientific perspective. You get a lot of very nice ideas from people from different backgrounds, whether it's cultural or scientific. I really enjoy working in interdisciplinary domains when we work on intersection of fields. Rather than just going deep into one field, I try to find similarities because very often in different domains, we are talking about the same thing using different jargon. - Wow, yeah. - Speaking English and Chinese, and we need to learn how to translate. There's also still my passion inside, like how can we build these interdisciplinary teams to build practical systems with the real-world impact? - And I think it's something important for students to hear 'cause it's a little artificial how in school you do biology for a year and then chemistry and physics, whereas actual problems have pieces of all of the disciplines. And that's sort of what makes them so interesting and challenging in a good way. - Totally agree. I myself struggled with this compartmentalization during the high school. I didn't really get it. Like, okay, why do I need this to learn calculus? I didn't really get it until I start working on scientific project. Oh, okay, this is how I actually applied and make it useful. And now, okay, now I use it every day. But back then in high school, okay, these are the rules. You need to learn it and that's it. Why do I need to learn it? That needs to be explained and demonstrated in more tangible example, I believe. - And my last question is, what advice do you have for high school students who are like science, maybe thinking about a career in science? - Follow your passion. Follow and nurture your passion because that's what will define your path. There are many paths in life. All of us have different initial conditions. And most of us don't have the same destination. So what I would say, what is important is the path. Whether you as an individual, you're enjoying the path for your life and have the progress that is your individual progress. Because especially these days, there is a lot of pressure, especially on young people, to perform, perform, perform, get better grades, get higher salary, do this, that, that. And it's very stressful. It's very, create this competitive environment and alienate people from each other. If you want to compete, compete against yourself, don't compare yourself with other people on social media. If you improve yourself from yesterday, I would say that's a great day. And don't overthink everything because lot in life comes from change, from change and random encounters. There is a chance involved. We cannot control everything. Universe is very complex system and we have limited senses and we limited processing power to compute only the small subset of what we actually perceive. So we can really think that we are interacting in a partial observable system and we are only computing the small subset of the information that we are getting every day. So there is this famous quote from very well-established computer scientists called Donal Knut. He wrote a very famous book, The Art of Computer Programming and there were a lot of foundational computer science topics. So he said, "Prematur optimization is the root of all evil." Don't try to overthink everything. - I love that. - Let's some things on chance and try to enjoy your life and follow the passion. And if you do that, focus on your own improvement. The things that is will not fall apart. Things will fall in line. Yeah, you will get where you want to get eventually. - Well, those are such great words to end our conversation with. I love that. I love that it came from a computer programmer 'cause I would maybe think that person would think everything should be optimized. Thank you so much for coming on the show. And is there anything you-- - No, thank you so much. It was pleasure. And yeah, thank you very much again for your invitation. - Thank you for listening to today's episode of Science Fair. Please rate and review the podcast on the podcast player of your choice. Also, please fill out a listener feedback form. You can find a link to the form in the show notes of this podcast or on the Science for Podcast website. Also linked to in the show notes. Finally, we are looking for episode sponsors. If you are interested in sponsoring an episode in exchange for us giving air time to your favorite cause, send an email to thesciencefairpodcast@gmail.com with the word sponsor in the subject line. This podcast is the work of me, Susan Keatley, and a fabulous team of interns. We have high school intern Lucy Poel, sound editing intern, Torin Gerbaz, and episode production intern Sierra Rebels.