Amazing topic and conversation! In this episode, Fatma Kocer talks with Justin Grammens about the fascinating work she and the team at Altair are by using Data Science and modeling to improve product design using structural optimization. Her quote that just being a Data Scientist, but needing to "know the domain" today, is spot on and very insightful. There are many other great quotes and insights from Fatma in this episode!
Fatma has received a Bachelor of Science degree in Civil Engineering from the Middle East Technical University in Ankara, Turkey, and an MSc and Ph.D. degrees from the University of Iowa in Structural Optimization. Currently, she works as the Vice President for Engineering Data Science at Altair. In this role, she and her team work on engineering data science strategy, development, and execution which includes investigating and applying the latest technologies in the field, providing feedback into Altair's software, and supporting customer projects.
Fatma is also a huge source of inspiration for women in the areas of STEM and education. as she is one of the recipients of Crain’s 2019 Notable Women in STEM recognition.
If you are interested in learning about how AI is being applied across multiple industries, be sure to join us at a future Applied AI Monthly meetup and help support us so we can make future Emerging Technologies North non-profit events!
Resources and Topics Mentioned in this Episode
Fatma Kocer 0:00
You need to be not just a data scientist, you know the domain you're getting the data to understanding data science. That doesn't work anymore. Maybe it did in the past, but now, if you want to provide efficient, effective data science and processing in an environment, you have to know the domain.
AI Announcer 0:23
Welcome to the conversations on Applied AI podcast where Justin Grammens and the team at emerging technologies North talk with experts in the fields of artificial intelligence and deep learning. In each episode, we cut through the hype and dive into how these technologies are being applied to real-world problems today. We hope that you find this episode educational and applicable to your industry and connect with us to learn more about our organization at applied ai.mn. Enjoy.
Justin Grammens 0:53
Welcome to the conversations on Applied AI podcast. Today on the program we are joined by Fatma koecher. Fatma has received a Bachelors of Science degree in civil engineering from the Middle East Technical University in Ankara, Turkey and Masters of Science and PhD degrees from the University of Iowa in structural optimization. She currently works as the vice president for engineering data science at Altair. In this role, she and her team work on engineering data science strategy, development and execution, which includes investigating and applying the latest technologies in the field, providing feedback and Altair software and supporting customer projects. pharma is one of the recipients of cranes 2019 notable women in STEM recognition. We also were lucky enough to have Fatma present at our applied AI meetup back in July 2020. So definitely check out that presentation at applied ai.mn. Thank you for joining us Fatma. Excellent, cool. Well, I gave the listeners a little bit of background and where you did your schooling and stuff like that. But kind of curious if you could fill in the dots, I guess maybe a little bit more what's going on today in your current role I mentioned Vice President of Engineering, what's what's sort of involved in that?
Fatma Kocer 1:59
Sure. As you also mentioned, my education and my background is on structural optimization. And alter, I wrote on what I explained, optimization tools are the most part of it. optimization methods and tools are very similar to what machine learning data science teams use. In fact, we also work on a lot of regression models for nonlinear physical phenomenon. So with this connection, about five years ago, I started looking more and more into new technologies in machine learning, and see how that can be applied to our domain, which is Computer Aided engineering, which I'll refer a C, which is a discipline that is helped in product design of many different industries, obviously, we located in Detroit, so starting with alternative aerospace, consumer products, electronics, etc. So currently, in my team, we are looking into how we can add value to cae in product design with data science. So we definitely don't want to do data science for the sake of data science, or you don't want to do it. Because it's the hot topic of today. We want to add value with it, we want to use it when it improves the process and the results of seeing. So there's a lot of methodologies out there that has worked for other industries, some happens to be very useful for our use cases as well. So that's what we're looking for is imagery imagery for what we're doing is probably closing the gap between data science developments and help engineering in engineering world uses it so we're trying to close that gap as well and leverage more and more of what's being developed in other industries are
Justin Grammens 3:57
very good, very good Sounds Sounds like a fascinating field. You talked about data science. And you know, this program, we talk about artificial intelligence, they obviously overlap, and they're very much related to kind of used interchangeably. It feels like in a lot different ways. But I don't know, do you have a textbook definition that you use for either of those?
Fatma Kocer 4:12
I understand that sort of simplification people use it interchangeably? I am of the opinion that we should not be using them interchangeably because they're two separate things. Yes, does use data science to build intelligence? designing an AI system is very different than designing and deploying a data science environment platform. So there are two very separate things. Ai does use data science, but I don't think it should be used interchangeably. My very simple definition is AI is a physical or virtual system that uses data science to take actions or for the decision process of taking actions and And their science is probably, you know, sort of the brain of AI. But again, the reason that I would like to separate them is because they're two separate domains, they have their own requirements, are you talking about of signs, you're talking about adding enough data training a model deployment model, understanding the results, when you're designing an AI system, you're looking into the human interaction, you can add, you know, the objective of the AI system, how it's going to be deployed, I was going to license it to the same things. an AI does use data science, but an AI is a physical virtual system that
Justin Grammens 5:40
Sure, sure, yeah, no, no, that's great to point that out, you know, I guess thinking about getting a degree in structural optimization, did you find yourself sort of learning a lot of the data science stuff along the way was this was a schooling that you did called a data science degree, I'm just kind of curious with regards to sort of how you got into the field. And if this has always been something that you've been passionate about, I guess, specifically what you're doing today.
Fatma Kocer 6:01
So that is a good point, I always knew that there is data science and optimization. At the end of the day, we use the same methodologies, you know, gradient descent, you know, an objective function with constraints, you know, we call it iterations, they call it the pulse, you know, you call it on paper function, we call it loss function, we use the same method, I was not until I attended a workshop on data science. And I left a workshop feeling like this with my structural optimization background, everything was similar, the terminology was different implications was different. And you know, the characteristics of data slightly varied from one discipline to another discipline. And at the end of it, you know, it's mad, it uses the same math, we are also looking into creating good data sets. And you look at the characteristics of the data sets to make sure that it's can be used for our purposes. regression models in optimization, because our attributes, we call them, our outputs, for example, are mostly continuous values. But you know, the transition from actual optimization to engineering data science is not as large as it may sound from our side is the same things, the math is the same, you start to see a little maybe slightly different. And you start also thinking few other things like model deployment, for example, in other environments, for example, one of the things that we start spending more time is about automated machine learning. So the transition is not that difficult.
Justin Grammens 7:50
Understood? Yeah, for sure. I mean, I guess if you think about the word optimization, and kind of how can you use math and data to optimize whatever it is that you're working with? So see, it feels like there's sort of a tie there? For sure. Right?
Fatma Kocer 8:05
As minimizing the loss?
Justin Grammens 8:09
Yeah, you know, it's funny, I guess, and maybe you've seen this, and it goes earlier this year, last year, something like that we're data science is the new sexy job that everybody wants to have. And my entire degree was in Applied Math. And you know, I'll show my age a little bit, but I was graduated in 96. And who would have ever thought that a mathematician would be a sexy job to have, but it's kind of come full circle, it feels like, you know, it feels like that. You're right, optimizing loss functions, all that stuff was, in some ways, at least back when I graduated, it felt like you would either teach or you'd go on to actuarial sciences, or you go on to a master's degree or PhD in math. And now it's feels to me like it's just like, you know, it's it is sort of coming full circle. Everyone wants data now that we can get data. You know, I think the Internet has changed that now. Now that we can get data and send data around, we need a lot more people that can actually analyze it.
Fatma Kocer 9:03
Maybe it's a controversial thing to say on my podcast. Data Science is analyzing data, right? The reason that there is a data scientist term is because there's no too small. There's no methodology and a separation was needed to be done between the data analysts of the past and the data. Age eventually was called data scientists. I always wants to be a scientist.
Justin Grammens 9:37
Yeah, yeah, for sure. But what's what's sort of a day in the life of a person who is set the vice president level doing what you do? Are you do you get into the data a lot or you spend a lot of time just managing teams and stuff like that.
Fatma Kocer 9:50
Don't get into the data and locked around getting to it to some extent. I like to get into it. I don't know if I have to know to be able to perform my Maybe sometimes it's actually an obstacle because you have to be a step away to be able to see the big picture to get in it. Because I, I like to read through the problems to the challenges when my team talks about them, I want to at least have an understanding of what are the intervals so that I can sort of feel the pain. To do few things, nothing, nothing super sophisticated. I definitely do not code or program notes and things like that. But you know, I do my own small studies on the site. Sure, I think it's essential for any level of people to have some connection to the work in the field, to understand what goes into that level of details, the level of complexities, the things that you need to think through so that you can appreciate the people that has done that work without exposing all the challenges to everyone. Yeah, we are a global company. So we start out the relative, you know, already with meetings, and the first thing I do is, you know, I check emails, despite all the warnings that you check in your emails, the first thing in the morning, you should do what's important for you. But also there is an environment that is open communication, and emails are not just information, but email is also our platform for discussions. And so it's important for me to read the emails first thing in the morning, to see what discussions are happening. And it also helps me to get to the mental stage that I was at when I left work the day before. So it's like my transition. And the record company, you know, we have a number of meetings, usually early in the mornings until noon ish, where we discuss ideas, discuss implementations, you know, plan on future projects. So those are reserved for our most internal conversations. And the afternoons are for myself and for customer meetings. And I would say three out of the five afternoons, I work on the products that the new release, and I work on testing them, making, trying to break them, trying to understand them, so I can talk to them better. And then the other days, I will do my own thing, my quick, small tests, or I read something or I refresh my memory on you know, normally I do this and that sort of thing. I like to working environment, and you should be doing your own tests while having a global idea. And also communicate globally so that everyone to know what's going on.
Justin Grammens 12:50
Excellent, excellent. Well, you mentioned products. And I think we've been talking about data and your background and stuff like that. I know, when you presented at the applied AI meetup, you get some great examples, I guess of some of the some of the interesting things that your products do. Is there anything you can talk about, you know, like publicly with regards to how you guys are using this, this data and a little bit more about Altair and their products, if you wanted to share that.
Fatma Kocer 13:16
In five countries out we have a very large portfolio of offerings from modeling and visualization for product design, to high performance computing to data analytics to IoT. And so data science is mostly working in our design modeling products. And our role goes towards embedding data science in our products to make them more AI systems. And our effort is twofold. One, we're looking into how to improve the process, and looking into how to improve the results the outcome of those processes. So for example, in our road in financial analysis, you have to get a design and as a senior analyst, you have to model that design, you have to mesh it, you have to apply the loads the boundary conditions, material parties, and that's, that's a manual process, it's a very detailed amount of effort, there's a lot of repetitive things that you do. So if you can reuse some of that repetitions and improve the processes that will not only improve the efficiency of the see animals, but it will also allow them to focus on more creative, more engaging work. So that's one place that we're working on. Another place that we're working on is remote with our optimization technologies. So you know, have descriptive analytics in order to improve a lot from different domains. And we're looking into how we can enhance those functions. So our main line of products in that domain is called hyper growth. But we also have other products, basically looking into how we can convert these manual processes to AI systems, AI systems that uses machine learning in the background to improve the process.
Justin Grammens 15:21
Okay, yeah, I need. I mean, like, if I were to be designing a physical product, I'd be using some sort of CAD tool, right. And so you guys integrate with the CAD tool to allow me to build a better product based on a lot of data that I'm getting in a virtual space.
Fatma Kocer 15:36
That's exactly what it is. So what we do is designer creates the CAD file for the application. And then in our products, you take that CAD file, and you idealize it, for virtual simulations to be able to run a finite element analysis, you have to create the finite element model from that CAD file. And that process is taking a CAD file, including a finite element model, is it pretty manual process, of course, we are doing a lot of automation at the same time, we can also look at automation, we're also enhancing it with data driven methods when when useful. So if we can shorten that process of taking a cab and converting to a finite element model, all the CNRS will last. That's what they want, they want to reduce that time as much as they can. Because that's where you do manual repetitive work, not learning anything, you just have to do that. So that you can get to the learning part, the learning part is when you have no model ready, and you run simulations, and from the simulations, you predict the performance of a part. And that prediction is based on, you know, the government physical laws. But that tradition takes a long time, because we follow a lot of differential equations. It's, you know, a fleet dynamics problem, for example, it's even more, it's a, you know, a crash application, like an alternative crash it takes, those are expensive simulations. The second part is really looking into how we can enhance them with data driven solutions, so that we can quickly do those quickly get to the performance predictions. And so the engineer can do more iterations, do more design exploration, and learn about the behavior of the physical system, and make a better design decision with all that collected knowledge. So instead of running two simulations, you can allow them to do, you know, 10 2030 iterations within the same time using data driven solutions with physics based predictions. And that gives them much better understanding of the physical system, and they can make better design decisions. So that's the second part that we're working on.
Justin Grammens 17:57
Yeah, I remember seeing a presentation must have been maybe a couple years ago, but I, and I forget the name of the company. But he went through an example where they were working with john deere, I guess, and they wanted to design a better tractor experience when you were sitting inside the tractor seat. And they modeled it all in virtual space, right? So based on pressure points in your body, based on things that you need to see outside the tractor, where should the mirrors be positioned? How should the seat be angled, all this type of stuff, and it was fascinating to me, because they would have had to, before all this stuff, probably make 50 seats. Or he would have to make a number of different types of things, and physically build it and adjust it all and bring humans in and have them sit down and go through all this stuff. And they were able to do this in a completely virtual space, which save them just a ton of money and time.
Fatma Kocer 18:48
Physical testing. conditions, using simulation was a significant improvement in the cost savings and in the time that it takes to do all these tests. And our goal is to take that one more step, which is combining physics based predictions using simulations with data driven solutions, and even counting the amount of time even more and more design inspiration.
Justin Grammens 19:19
Sure, sure. And that's what the designers want to do that designers want to design.
Fatma Kocer 19:23
Exactly. So we won't have to go through that painful modeling process. Learning you are learning about possible ways to improve the design. Are you testing those out? That's actually the fun part of engineers and designers work?
Justin Grammens 19:44
Yeah, yeah. And people have have this fear. I guess that my job is gonna be automated away. It feels like what you guys are working on and your your tool sets are actually making people be more efficient and more effective. These are very important tools.
Fatma Kocer 20:00
don't relate to the question of my job will be taken away in this domain. Because the minute we can get rid of these new sort of energy draining tasks on your day, the moment you can actually get to the engaging part of your work, and that's not removing your job, but that's actually improving your job.
Justin Grammens 20:20
Any advice for people maybe entering the field? Any classes or things that you've seen? I mean, obviously, you followed a certain trajectory, by I mean, I guess even if you step back, who are you looking for when you try and hire people? What are some skill sets that people should have if they want to get into this?
Fatma Kocer 20:37
Different than data signs in other fields, for example, you look at classes, you're not going to see any engineering data science example. But you most likely see, like recommendation systems, examples from healthcare examples from financial institutions or marketing. And other fitness mentioned at the beginning. We as engine data scientists, one of our responsibility is to close that gap. So because the field is very new, it's not as mature. This is due to some of the challenges that we're tackling in working in data science. So there's no one class or one program that I could suggest that when I was shifting my focus from optimization design inspiration to data science, the first thing I did is look into some of the courses that are available in Coursera. For example, I really liked him because they were refreshers, widen my horizon and topics that I didn't know, I read a few conferences that were eye opening, there's nothing that I could take my word for it, there was plenty of CO working teams that I saw in those conferences. If you want to get into engineering data science field, I imagine this is actually true in other fields as well. First and foremost, you have to develop an expertise in the field, you need to be not just a data scientist, you no longer can say I just do data science, you're getting the data, I'm just going to do data science, that doesn't work, maybe did at the very beginning. But not anymore, because there was a separation of jobs. And we were at the beginnings of applying data science. Now. If you want to provide efficient, effective data science tools and processes in an environment, you have to know the domain, there is no ifs and buts to it. If you want to be an engineering data scientists, you have to know the engineering domain, you have to know what the challenges are, you know, you have to know what you need to work on. So interestingly, for us, I always say, opportunity, challenges. And those challenges became our opportunity. To solve these challenges, we said, okay, well, we can work on solutions to overcome those challenges. And those became our opportunity. So you definitely have to not only focus on data science, but also pick a domain and be an expert in that domain. And I also like to attend conferences, or listen to other people talking on this topic, because they bring momentum to my thought process my eyes to other things. So I would definitely encourage people to attend conferences, I know some of them can be financially and timewise may not always be very attractive. But I think it's important in the discussion about how people in Europe see the value in attending conferences and sharing this information, but maybe not as much North America. But I definitely suggest people to attend events in your field when they feel like they're not being as creative in the solutions that you're bringing, because seeing somebody else working on something else makes you
Justin Grammens 24:00
very good. Very good. Yeah. I mean, I'm assuming there's there's certain conferences that are specifically focused on engineering data science versus just broad ones. Is that true?
Fatma Kocer 24:10
It's not yet a conference that is that is focusing on entering data science, but there is a number of engineering conferences that has data science branches or sections in it chosen a fence, which is very focused on numerical analysis, finite element methods. So they have initiatives that are really nice. So that's one reason I once went to an intermediate conference, which very little to do with what I do, but it was an eye opening just to energy, things that people are doing in other domains. It was one of the turning points for me.
Justin Grammens 24:48
Very good. I mean, I I said during the beginning that you received the cranes 2019 notable women in STEM recognition and I'm just sort of thinking about the field in general. There is a lack of females in this field. Would you agree?
Fatma Kocer 25:00
There is definitely a lack of females in engineering. Data science in general is a little better. I think there's there's more in data science, but probably not enough. I mean, the sad thing is, I'm not sure if this is still the current situation. Nowadays, there's a lot of females that graduated from engineering and data science. And it doesn't translate into the work environment. For more females that goes in into engineering, graduates from engineering, they need to start working after a couple of years they drop off. And that I think that is the sad part, right? Well, in the past, it was that they didn't go to engineering, I should say, we didn't start dropping off from the workforce. And, you know, there's a lot of advices on it, you know, like, one advice was, well, you know, maybe we should be more like men. And it should mean in, I'm often the opposite opinion, I think everyone should be able to are and but everyone should be accepting of their behavior. And just like, you know, welcome people that hope and work differently, because we see a lot of very successful engineers and data scientists.
Justin Grammens 26:20
Yeah, for sure, I think getting a much more broad perspective. And I even would say, if there's anybody out there that's listening to this, that that is a woman that would like to be on this podcast in the future, I would love to talk with you. Because Yeah, I think, you know, we've done by the time this airs, we'll probably have done at least a dozen or so. And you are the first woman actually, to actually be on this. And I don't want it to be that way, I really want to have a lot more women. So if anybody has any recommendations, you included, please point them my way, because that's what this is all about is actually getting everyone's perspective and, and talking about different applications with data and artificial intelligence. And you can't design systems that serve everyone's need, if it's only designed by a certain group.
Fatma Kocer 27:04
Exactly. Sure, I have multiple names in my mind right away. And I will definitely make those suggestions. I'm honored to be part
of it, I understand why this happens. The future looks promising. So I think it's more and more
Justin Grammens 27:22
good. Well, how do people reach out and contact you?
Fatma Kocer 27:26
If anyone works in this related field, I'll be more than happy to have discussions. You know, I had interesting discussions with people from different fields. And at least the people that I reached out are people that has worked in relevant topics, or in topics that I'm interested in, for example, the philosophy of AI, you know, and how it impacts the social environment is something that's interesting to me, besides, you know, engineering, like where should the focus be in AI development? Should governments have any oversight over? It just just reach out to people? Are you a little people that says, Well, I don't want to bother you, but it's not bother. I mean, these are all interesting, new things, discussions is, is always good. So it's never a bother to have someone reach out to me and start a discussion on a topic that is that I'm working on or that I'm interested in. And I can I would repeat this again, to the reader after it's never better to, you know, if you reach out to someone to start a discussion, I think there's the momentum of those discussions are what keeps the room going, I have someone that instead of in my mind, and I get productive only after I discuss it with people get some perspective to it. And then so I can fine tune what I'm working on, and be successful. So I'll reach out to anyone in the industry and anyone in academics that's working on relevant topics, and you know, sometimes shy away from it as well. And if I don't hear that's fine, you know, it's not personal. And when the time's right here, it's always really, really great conversations.
Justin Grammens 29:11
That's great advice. Great advice. Yeah. So we've covered a fair amount of ground here. Is there anything else that you maybe would want to talk about?
Fatma Kocer 29:20
One of our past discussions, I think you open the topical with how the working minds or work is gonna change with all the developments in AI. I think we mentioned that as in, you know, AI is not gonna replace or it's gonna improve. Um, there's one line of code that I know now that you asked I want to be which is the removal of this routine road and getting us to do more engaging, more creative, we're not going to be working from eight to five anymore. We're going to be working 24 hours, but it's not going to look like work. I think you're going to be working when you're walking your dog or when You're actually most of these creative thoughts. It comes when you're not actively working on you just after doing some random things. So I tend to work environments, you know, the managerial skill sets would have to adapt to it. You're not going to be guiding someone that's sitting on your desk eight to five, you're gonna start valuing people that is coming up with ideas that's taking risks that's failing. That's learning and that's communicating that well, that's the moment that you know that we have it out there. But that's also the mind that I'd like to promote elsewhere. Because I think that's what's going to be useful. Engaging, interactive, include word. Absolutely,
Justin Grammens 30:41
yeah. As you were talking about that there's a book out there called, I think it's called a whole new mind, by Daniel Pink. Are you familiar with that one? No, I'm
Fatma Kocer 30:49
Justin Grammens 30:50
Yeah, it's really good. And it really talks about how at least, you know, in the 50s, and 60s and 70s, kind of everyone's pushing the left brain logical thinking, right? Everyone goes to school, we learn our equations or math. And we learn history, there's these very, very concrete things that we learn. And the creatives and the art stuff is kind of left to the side, right? Most people don't go into those fields, because it's really not pushed hard in school, and his whole philosophy. And basically, his prediction is there's a revolution right now going on, it's been going on for a decade around right brain thinking. And around, it's the creative thinkers that are actually going to be pushing the next generation of workers. And these are the types of people that actually are going to become valuable. It's not going to be I mean, obviously, left brain thinking and science and all that stuff has its has its place. But his thing is, is that basically the right brain has been discounted for too long. And we should start exercising that part of our well being and our work is going to change because of it.
AI Announcer 31:50
Well, no, I
Fatma Kocer 31:50
definitely agree with that. I'm going to check out that book. I'm not the most creative person, I think.
Justin Grammens 32:00
Fatma Kocer 32:01
Because there's a difference between one person to another person would be how they can meet you. knowledge that you've gained and that they are looking at challenges.
Justin Grammens 32:16
Well, great, I got a good future ahead with all this cool technology and all this creativity going on for sure. Our industries are going to be changing and evolving, which makes it exciting. Well, great pharma. I appreciate your time today. Thank you so much for being on the program, and I look forward to keeping in touch in the future.
Fatma Kocer 32:34
Well, thank you, Justin, thank you so much for the opportunity
Justin Grammens 32:37
AI Announcer 32:40
You've listened to another episode of the conversations on applied AI podcast. We hope you are eager to learn more about applying artificial intelligence and deep learning within your organization. You can visit us at applied ai.mn to keep up to date on our events and connect with our amazing community. Please don't hesitate to reach out to Justin at applied ai.mn if you're interested in participating in a future episode. Thank you for listening