How Ivy Leagues Teach Data Science | College Professors Share All

And let's dive straight in. So I think first of all, it'd be nice to know a little bit about the programs that you the data programs that you have at universities. So, Joel, do you want to kick it off? Like, what does the data program at Git involve?

Sure.

So we actually have a lot of different programs. So we have our flagship Master of Science and analytics degree, which is an interdisciplinary degree hybrid between data science and analytics. And we offer that degree in person here in Atlanta and fully online from anywhere in the world. We also offer a bunch of related degrees from the undergraduate level through the PhD in fields like computer science, machine learning, business analytics, statistics, industrial engineering, operations research, and more. And then we've also got some shorter certificates including analytics and data science and business analytics.

Okay, so I guess that shows how data spreading around all over the place. You can do not just data, but you can also do other degrees that involve a bit of data as well. Okay, nice. Joe, how about wgu, one of the programs like that?

Similar to Joel, we have a portfolio of offerings as well. Within the undergrad space. We have a data analytics bachelor's degree that actually just had a pretty significant relaunch back in September, and we can talk a little bit more about that later. We also have a master's in data analytics. We have an undergrad in computer science, software engineering, cybersecurity, network engineering. You know, a lot of it is really practitioner oriented. It's all about what can we do to get you a job in the tech sector as fast as we possibly can.

I like that. So there's a definite end goal there. And maybe we'll get into that in a moment. First, I'd like to talk about who the learners are. Like who's actually wanting to go through a degree in data maybe this time. Joe, do you want to go first?

Sure, be happy to. What's kind of interesting is analytics, and that's really where a lot of my passion area is, where I've spent a lot of time in industry. It appeals to people from such diverse backgrounds and experiences. I can remember having folks as diverse as a firefighter, a nutritionist that came into data science and folks that are coming from your non technical, non STEM degree kind of backgrounds. And they really find it interesting because of this blend of the creativity, the soft skills, the storytelling, the ability to solve problems and then use data to be able to effect that change. So meaning to create change, which has actually created an interesting opportunity within the academic setting because you're now trying to build these programs for Folks that may have been born into data and love to program till 3 o' clock in the morning as well as folks that this may actually be their first adventure into the data space and want to have that career transition into being a data analyst, eventually a data scientist, maybe a machine learning engineer. So I'd say right now in our population it's about 50, 50, about 50% of folks are coming from non stem, non tech backgrounds directly into our programs, and about half are actually coming from prior experience in data, prior experience with math, computer science, etc.

That is absolutely fascinating. Like just the, the two sort of extremes there, like people who are like very hardcore STEM folk and then other people who just got like no experience come across data and some of the technical topics for the first time. Joel, have you seen a similar thing?

Yeah, so, so we have, we have a lot of people who have the sorts of backgrounds you might expect in like engineering, computer science, stats, math, economics, business, those, those kind of things. And then we also have a large group of students who have, have degrees and professional experience in things like dance and theater and law and medicine and archaeology, sports, astrophysics. And so we've also had to really think about how to structure our programs so that they're accessible to students who have just a small number of prerequisites and take them all the way through everything they need to be a successful professional. But also if you're coming in with more experience and more background, how to let those folks take advantage and maybe take some extra electives instead of taking those, those core background courses that they already know. We've also noticed a wide range in age. We've had students just in our master's degree program from age 16 to 84 and we've got them from all seven continents. We even had a student in Antarctica. And so it really is sort of giant age range, giant geographic range, and really large background range.

That is pretty impressive, I have to say. Well, both at both extremes of the age range, like being able to take a degree at 16, but also taking a degree at 84, that's, that's some commitment there. Excellent. So I'm guessing the 84 year old wasn't taking the degree in order to get a job. I think like maybe a different motivation there, do it more for fun. So on that note, I'd like to talk a bit more about like the motivation. So I guess some people are taking this to get a job. Have you seen any other sort of reasons for taking a degree in data?

Yeah.

Joel, do you want to Go first this time.

Yeah, so we, we do have like I, I Talked to the 84 year old, he had worked on the Apollo program back in the 60s and just wanted to keep his mind fresh. But most of our students really are, are looking for a job in one way or another and it's split about, about one third are already in data science and analytics jobs, but they realize that they need more training to do those jobs well or to move up and so they're looking for that sort of enhancement. Another third are people who have jobs in some other field and they want to transition to data science and analytics. And then we also find about a third are people who have jobs in some other field and, and just want the data skills to supplement what they have already. So a doctor who wants to use data science in his medical research, for example, those sorts of things. So we really get big populations with all three reasons and we try to cater to all three.

Excellent. Yeah. So it's like I want a job in data or I'm doing a different job and just need some data skills to support that. I like that. Joe, have you seen any of the sort of motivations for doing a data degree?

You know, I think Joel hit a lot of them. I think the one group that may also be in there are folks that may have learned on the job, they may have been using different webinars, moots, different things that are out there and basically became self taught and they're now looking for some kind of a credential to either move up in their current position, be able to get a new role, or just be able to go in and figure out what are the general practices and frameworks that maybe they didn't learn on their own that could help them to really advance in their profession. So that's one of the reasons our competency based education model really comes in. Because if folks actually have skills in SQL, in Python, in R, in stats, they can really breeze through a lot of this content quickly. They can be able to show that they have that competency through the assessment and then move on. So totally agree with Joel. I think you have a lot of these career advancers, you have some career changers, but I also have, I know quite a few people that really learned themselves. They figured it out, they went onto Reddit, they went on the Stack Overflow and they were playing in GitHub and they just wanted to be able to get that framework around it and to be able to get that hard credential.

That's interesting. It's quite Reassuring that a lot of the stuff you're talking about is very similar to the experience of Data Camp users. It's like people wanting jobs, people wanting some credential to progress their career. Excellent. Let's get into the meat, like what you should be learning about. So maybe to start with, let's talk about what are the main sort of data topics that get covered within your master's degrees? Joe, do you want to go?

Sorry, did you say Joe or Joel? I know it's gonna be fun.

Yeah, I'm gonna say top mumbling. Yeah, Joe, you go first this time.

One of the. One of the reasons I came into academia from the corporate space was to be able to rethink analytics education a little bit. Especially because a lot of the programs going back a few years were extremely technical. They were hard data, hard math, hard on the computer science side. But those business influence skills, they may have been alluded to in a few places, but they were never an equal partner in that mix. So when we think about our analytics curriculums in the master's program and in the undergrad, it's really a balance of three levers that we look at. One around data and programming, one around mathematics, and then one around business influence. And the program actually represents a full third in each of those categories. So they are truly equal partners in that mix. And when I start diving into the data piece in particular, there's a lot of emphasis on foundational data management. Especially now with the ChatGPT and the large language model evolutions, everyone's realizing that if you don't have a great eye on your data management practices, it's garbage in, garbage out. And that's becoming very evident very quickly, especially now as the Gartner hype cycle is starting to turn at the top of its peak a little bit on the math side, it's also building those foundations. And to the point that Joel and I were talking about earlier, when you have some of these students that are coming in that may not have really been in a STEM tech background before, we want to make sure they have the readiness and scaffolding to be able to succeed, Especially if they're diving right into a master's program, they're going to have peers in there that may have been spending a lot more time in that space, and we want to make sure they have the additional resources to succeed. So data management foundations, building the foundations within the stats and the linear algebra and everything else you would need from that foundational perspective. And then on the business influence side, we prioritize storytelling, we prioritize creativity and design thinking. We're starting to really think about decision intelligence and emerging discipline as well, and how that's going to play in. So the combination of process people and technology and how those three things actually enable a data scientist to really create changes in an organization. So going beyond just creating the algorithm. So that is the balance of those three levers. And then we opt to give our students the ability to move some of those levers up and down. So if they are really a data nerd like I am, and they want to go deep into that, they can, but they're still going to get what they need from the math and from the business influence to be successful in our role.

That's fascinating, the idea that you've got these different topic areas. So the business of the tech stuff, the statistics, data management, and you can just tweak it a little bit depending on what your goals are. Okay, Joel, do you. How does your program work? What are the different areas?

So it's actually. It sounds pretty similar from a philosophical point of view where I mentioned earlier, we have an interdisciplinary master's program, and the three units are the College of Computing, the College of Engineering, which has most of the statistical and operations research modeling pieces, and then the Scheller College of Business. And so our core covers all three of those areas, both as individual perspectives and also in an integrated way that helps people sort of learn how to put those three perspectives together. And then the other half of our courses are elective, and so students can choose to specialize in any of those areas. On campus, we've got more than 100 elective choices. Online, we've got 20 or 30, I think, elective choices. And so people can really sort of start with that core and then customize the program to whatever. Whatever their interests and career goals are.

Okay. So I like the idea that you got some core skills that you need and then, you know, you sort of choose the rest. I have a sneaking suspicion a lot of today is going to be all three of us agreeing on what the important things are. Certainly like is both those things are very similar to how the data camp certifications for being a data analyst or data scientist work. It's like, yep, you need some technical skills, some business skills and. Yes, and data management skills. So actually, some of these things are very, very different. So learning how to do data storytelling is very different from, like, I'm coding something in Pandas. Like, particularly like the deep learning stuff, it can get quite involved. Is there a natural order to.

This.

Which, which things you should learn. So which skills do you need first and which ones come later on? Joel, could you start this?

Yeah.

So I, I think you really need the fundamentals first in all three of those areas. So we like, we've got some fundamental courses, one in, in the, the data and modeling and, and, and that sort of side of things. And then we have another course in pipelining and coding and scraping and all the data sorts of aspects. And then another course in the business application, business fundamentals, business application of data science and analytics. And then once you've got that, you can really branch out and learn things in different orders. But it's tough to really, like you mentioned storytelling, it's tough to get a good handle on storytelling. If you don't know what the goals of the business are, or if you don't know what the capabilities of the models are or what the limitations of the data might be, then it's really tough to start telling those good stories. So we really start with those fundamentals and then really build on top of that. Deeper analysis methods, visualization, deeper business applications, more complex data science methodology, et cetera. They all kind of sit on top of those fundamentals. And then in our program, our last course for everybody is an applied practicum where everybody has to go out and do a significant data science or analytics project with some company or organization that really cares about the result right now. So there's never a recycled project or a toy project or anything. They're all working with these companies right now on something to get them a semester's worth of experience before they graduate and go out to work.

Excellent. I do like the idea that you're not just taking courses. You are then going and doing some projects and you're applying your skills afterwards. Joe, do you want to elaborate on what constitutes fundamentals and what order you need to learn things?

Yeah. A lot of it comes down to scaffolding and the way that Joel described it, starting with the foundations is definitely something we believe as well. The one thing I will say though is this kind of goes back into the left brain, right brain debate and the fact that a lot of folks in the advanced analytics space may feel uncomfortable with either the technical or the soft skills. In fact, the true unicorn almost becomes the person that is comfortable with both of those things. But the one thing that I'd like everyone watching to really appreciate is the fact that you can do either with the right scaffolding and the right practice. So I used to be completely stage fright every time I Go into a meeting, I would talk about the regression outputs, I'd be all excited about the code. I even had a math tie at one point. I was really uncomfortable with storytelling. Some of the things I did was going to get into improv. I actually took improv lessons. I became a bartender on the weekends to figure out how to talk to people. A lot of those things you can do in parallel as you're learning from some of the coursework as well. If you think storytelling is going to be your Achilles heel, you may choose to supplement some of that in your progression. And doing that early on because as you start to think about, okay, well, how do these data management foundations and the math foundations all come together? You're going to realize that storytelling becomes really important there. In fact, one of the first exercises I used to give to data scientists that were struggling with it is how do you tell a story where there's not one data point actually in your story and actually getting to what the business user really is after. Now if you are a master storyteller and you're really comfortable with people, well, maybe it's on the other side. Then you're going to be spending a lot more time practicing on your programming, thinking about algorithmic thinking up front before you actually step foot in SQL, Python, R, Java, whatever it is that you're playing in algorithmic thinking, getting a whiteboard, being able to visualize what you're trying to build regardless of the coding language and then continuing to build from there. So figure out and do a little self reflection on what are you comfortable with and where do you think you're going to struggle and see which of those two camps ends up leading you on and then step into it that way.

Some great tips there. Because you mentioned that you used to do improv classes. That seems to be a surprisingly common thing with people in the data space, that they've improved their storytelling skills either by doing improv or stand up comedy. I used to go to a storytelling night where it was like, you know, turn up to a pub and read stories to other people there. So yeah, there's a great tip for getting better at that sort of thing.

Makes for really entertaining team meetings too.

Absolutely Excellent. Joel, do you have any background of that? Have you done any sort of weird things like improv or whatever for practicing speaking?

I haven't, but. But just getting up and being a professor for 25 years, taking impromptu questions from all your students kind of gets you into that, that mode too.

Yeah, job training.

I may be less funny, but. But the skills are nice.

Okay, so we talked about the sort of the having core skills that everyone needs and then maybe customizing it depending on what your goals are. So I think like maybe the three most common job titles at the moment in data are going to be data analyst, data scientist, data engineer, and then there's like 50 other related job titles. Is there a difference between what skills you need depending on the job title you're sort of working towards? Joe, do you want to go first this time?

Sure. And, you know, I'm actually going to use that lever kind of metaphor that I was using earlier. So first off, whenever you're going to look at any of these job descriptions, really get into the fine print of what the company is after. Because what is a data analyst at one company, maybe a data scientist at another, maybe a machine learning ops engineer at another. So look for the keywords. But the traditional data analyst, the way I think about it, is actually a balance of the three levers. I said earlier, they have a foundational understanding of data, foundational understanding of the math, and a foundational understanding of the business. From there, they may choose that some of those levers are going to go up or down. So they may choose, okay, data engineer. I'm going to move that data lever all the way up. I'm going to start thinking about the data ops space. I'm going to get up to speed with what a data lake house is and, and how it's kind of a nod to a DBA coming back into the loop. In a lot of ways it's funny how history repeats itself. Or you're going to realize, you know, I want to do, I want to be a statistician, I want to just play on the math, I want to do controlled experiments. Or you may decide that you want to be the unicorn of a data scientist, where those levers are going all the way up. So I think a lot of it comes back to preference. You know, what are you good at? What do you enjoy? And then how can you balance those three levers to be able to create a role that really makes sense for you? But understanding that a great way to step into these teams. And I'm putting my, my Dell had my EMC back on is coming in as a data analyst, you know, get enough exposure of the three, you're not going to be successful without that. And then figure out what really speaks to you and what drives your passion. And then you can take additional trainings, additional courses on the job, of course, and you can figure out which of Those jobs you're going to ultimately land on where you're getting on a bed in the morning and you're really excited.

Excellent. So, yeah, I really do like that idea of leave us to customize things. Joel, how do you divide up the sort of the content? So different skills for different roles.

Yeah. So I agree with both of Joe's main points. First of all, there's not really such standardization in the HR space yet. And so what's called the data scientist at one company may not be the same as what's called the data scientist at another company. You really have to look at the job that you're doing, you're applying for and fit that profile. And. And I agree too. I mean, our. It sounds like both of our programs are really designed to let you pick what you want to specialize in and, and go ahead and do that. But I think it's. It's also, I just want to add that, that even within a category, even if you think about sort of the traditional, like a well defined data scientist category, one data scientist job might require strengths in time series and Bayesian statistics, if they're in one field. And another industry might require strong skills and deep learning and optimization, and something else might require natural language and experimental design. And they're all called data science. But because of the industry and the part of the business they're in, they require these different skills. And so we're finding that employers are really digging deeper than, than just what does the credential say to see exactly what skills you've got and whether you're a good fit for these jobs.

Okay, so that sounds like you might need to do some reading of job adverts if there's a particular industry you're interested in, just to see what are people actually learning about.

Yeah, and. And the other way, the other direction too, that, that just saying I want to be a data scientist might not quite be enough. But if you want to be a data scientist in the finance industry, for example, go see what they're looking for and, and really specialize in those too.

Brilliant timing because we had Data Frame podcast episode that came out on Monday where there was someone from an investment bank talking about how to get a job in algorithmic trading. So, yeah, maybe Reese can post a link to, to that because. Yeah, that's. That's worth a listen. All right, so we talked a. About technical skills and what order to learn them. We also touched a bit on soft skills and how you need storytelling and a bit of business knowledge. I'd Love to go into that a little bit deeper. Joe, can you talk me through, like, what are the soft skills you need to work in data?

You know, the first thing that comes to mind, I've been having a lot of conversations about this recently, especially with everyone figuring out how do I implement genai and large language models? It's an appreciation for the ecosystem in which you're working and that ecosystem includes people, process and technology. So when I start thinking about soft skills, it's being able to do process analysis and mining. It's being able to actually go sit down and have a conversation with somebody who's going to be using whatever it is that you're working on, doing those interviewing, building those relationships, building trust. So really bolstering your EQ so your empathetic skill sets and figuring out how you can really embrace that and really use things like empathy maps, trying to understand what, you know, what somebody says, does, thinks and believes and how that's going to balance into the effectiveness of whatever it is you're trying to create. So it's really got a design element, it's got a very human touch perspective. And again, realizing that the data, the math, the algorithm is just a piece of what you're going to be doing, at the end of the day, you may end up creating the most amazing algorithm. You may be using the newest technologies. You figured out how to do a large language model, internal, within your enterprise solution. You were able to put that on top of something else you were doing. You're all excited, you go into a meeting and the business is going to look at you and say, okay, now what? What does that have to do with the way that we work? Do you need to change a business process because of that? And again, some of this starts pulling back on storytelling. So the way that we approach it, especially in the undergrad, is building that foundation around, okay, design, thinking, storytelling and an appreciation for the process in which a business works in that ecosystem. And if you can really get your hands around those things and then look to find ways to bolster your emotional intelligence, you're going to be in a really great position to really drive ROI with the products that your greater data science team is starting to create.

So some of these things actually sound pretty difficult to acquire. Things like empathy is like, you know, it's going to take a lot of practice. So I always feel like soft skills is kind of an ironic name because these things are hard to acquire.

Yeah, well, no, it's interesting. You said I had a coach, a career coach at one Point that, you know, I think I was doing a. One of the Hogan360 assessments, which, if you're not familiar with that, it basically tries to measure your EQ and your emotional intelligence. And it's really something that you're born with in a lot of ways. There's a nature and nurture component to it, but there are mechanisms that are out there and resources to help you to fill the areas that you may not be as strong in, or if you are overly emotional, intelligent to the. You may not be pragmatic anymore. How do you walk it back a little bit too? Sometimes emotion can actually get in the way of being able to operationalize and execute. So if some of these things are starting to sound a little scary and that, hey, you know, I came here to program and to work in data and build models and be able to get my R squared up and all these things, you can do this. I've seen people that were highly technical become these people. I was highly technical, and now I'm here talking to you all. I love talking to people. As long as you have the right coach and you have that right foundation and you go into it with that mindset that the math and the data is not the be all, end all in actually driving roi.

Okay, I like that. Yeah. If you want to have an impact, you need to be able to persuade other people or explain to other people what you're doing. Okay, Joel, are there any other soft skills that you think are important in the world of data?

Yeah, so I, I guess there are. There may be three that I would add. And this is where I sort of get up on a soapbox sometimes, so I apologize in advance, but I think. So the sort of. The storytelling piece is, is obviously important and that understanding of where you are, but I think we also need to be good communicators on the incoming side. Being good listeners and, and being able to ask questions and then listen and internalize those answers, I think is really important because often people in this data space are called on to, to work in areas that we're not. We're not trained in, we're not comfortable with. We're being brought in as a sort of an outside professional to help in. In some area of the business. And so we have to be able to gain that understanding. And so we have to be good listeners and good questioners. I think also our field is, is moving so quickly and advancing so quickly. I've heard people say we have to be curious and, and that's certainly true, but we also Sort of have to be professional learners that whatever any of our programs are teaching now. We'll teach you what's state of the art now on the technical side. But in five years, there's going to be new stuff that just didn't exist when you were in our programs. And so we try to teach people how to learn, how to learn essentially how to pick up new things quickly so you can stay ahead of the curve as you're going through your career. And then the third one, which is really sort of my soapbox thing, is that, is that all these data fields seem to be. They're, they're kind of professional fields like medicine or law or plumbing or auto repair or anything where we don't know. And we go into a professional and say, hey, you're the expert. Tell me what needs to get done and go do it. And we just trust those professionals, whether it's, it's my doctor or the guy who fixes my car or anything else, we trust them to tell us the truth because they're professionals with this specialized knowledge. And we're really in that position too on the data side. And so that means we can't let our preferences or our biases affect the outcome or the reporting of what our analysis says. We need to really be true to our analyses and true to our profession. And that's something that we try to instill in our students, too, that, that this isn't just another field. This is really a profession and you have to treat it that way.

Absolutely. So, yeah, all important stuff there. I do like the idea that you need to have some kind of integrity there and not be like lying with statistics or gamblized, I think is the traditional phrase. Yeah, I also like the idea that. Well, maybe I don't like the idea, but I do feel like it's true that you can't just get a qualification and be done with data. You are going to have to continually learn throughout your professional career. Like, I feel like I've been, I've been working day for 20 years and I'm like, yeah, hopefully I will know data science one day. I still feel like I'm a long way off from that. All right, so the other area where things are changing pretty rapidly is generative AI. Obviously it's become wildly popular in the last year. So are there any particular skills that data people should have in this area? Joel, do you want to go first this time?

So right now I think the biggest skill that people have, unless you're sort of, you want the sort of job where you're on the forefront of cutting edge Gen AI. Other than that, I think the biggest skill people need right now is to understand what Gen AI really is and what its limitations are. There's so much hype around it right now and so many companies and organizations are just wanting to use it for everything, throw Gen AI at everything that we, we really need knowledgeable people to be able to put the brakes on in some areas and say, hey, this, this tool isn't going to work for this problem or, or even worse, it might be counterproductive. So I think that's, that's really important right now. And then I think another skill is just how to use it effectively. It's, it's like any other tool. Anyone could pick it up and try to use it, but a little training really goes a long way if, if you're trying to help someone use it more effectively. And I think that's really a key piece right now too.

Absolutely. So I like the idea of understanding the limitations and like what's possible, but also what's not possible. Like yeah, probably don't try and write a whole PhD thesis. It's not going to give you good results. Joe, do you have anything to add to that? Like what do people need to know about generative AI?

You know, one of the keywords that Joel said there is that it is a tool at the end of the day and there are some folks right now with the hype that treat it as the be all end all at the end of the day. You got to think about how is it that my business is going to use this? What does it mean for the way people work? There's likely a partnership that you need to have with your HR function as well too. Maybe you're redefining job profiles and what the actual day to day tasks are and to the limitations to the point you're talking about that starts talking about foundational data management again, security. You don't want to just be going into generic OpenAI and putting something confidential in there. You want to start thinking about gatekeeper technologies to be able to make sure that something that's confidential and proprietary isn't going off there into the World Wide web. I think where this is a little bit different is in the past when people were playing with R or Python or SAS or spss, you typically had two kinds of deployments. You had it on your local machine, which may turn your motherboard into a heater depending on the size of the data set. You're trying to play with or you're doing it in some kind of a private cloud environment. In both cases, you didn't really have a lot of those risks from a security perspective because you typically had it ready to help you sit in that cloud environment or are you doing it on your local PC or Mac or whatever it is. Now with a lot of this new advancement, you have all of these different applets that are popping up on the Internet that make it easy to digest data, get an instantaneous response, and don't have a lot of those security considerations that honestly a lot of us just took for granted in the world that we used to be in. So one of the things we're trying to emphasize with all of our students right now, and that's not just in the college of it, but across wgu, is limitations. What assumptions should you be making and not making? And also trying to differentiate between effective use, to your point, Richie, versus trying to build these algorithms from scratch. We have some folks that I talk to, prospective students that they want to build the next ChatGPT. And you know, my conversation with them is more around, okay, well, These are the PhD programs you should probably be considering.

Absolutely. There's a vast difference between I can write a decent prompt for Chat GPT and I can build Chat GPT. It's like a lot of years of working with PY in between that. Okay, all right. So I guess related to this because Chat GPT is actually pretty good for learning. Like for example, on Data Camp, we have tutors built into our GPT based tutor built into our courses. We got GPT built into workspace. So if you want to write some code, it can help you out that I'm presuming this is going to have a big effect on how people learn at university as well. So maybe, Joel, do you want to talk about how you can use Chat GPT for learning? What are your students doing with it?

Yeah, so I think there's, there's really a wide range of what people are doing with, with Chat GPT in terms of learning, some of it more effective than others. But I think, I think really where it's going will be sort of more the personalized tutoring. I think it's easy to ask the question and I've seen different levels of success regarding how well your question actually will be understood and an answer returned. But there are some really good ones that, that I've seen out there. And, and so I think that's really the way that it's going. We have like, both our universities have, have some Online programs where you could get help 24 hours a day, basically. And, and so I, like, in my course, on average, I think questions are being answered every 12 minutes right now. If we had an AI that could answer all those questions, it could be every 12 milliseconds that a question is getting answered. I think it can really help learning in that sense. But there's still the buyer beware aspect, that these language models are trained on language rather than on technical material. The answer you get isn't always the correct answer. And so students have sort of been learning that they need to be a little bit careful about just taking the answer and repeating it without thinking through, is this answer really correct or really what I'm looking for?

Yeah, I do like the idea of having a chatbot, like maybe a personal chatbot, just to help you with all your questions. And I guess from the professor point of view, not having to deal with, like, lots of student queries is probably great for the job as well. Okay, Joe, do you have anything to add to that? Like, how are people, how are your students using chat GPT?

You know, I think the first thing we all have to acknowledge is the fact that when you get into a professional role, you probably will be using some form of chatgpt in the future state. It's not, you know, it's kind of like the calculator, when that first came out and you had math teachers picketing and there was some really fun images from newspapers that were popping up when OpenAI had their announcement last year. Because of that, we need to think about what it means for the way we work now from an academic perspective, I'm thinking about two things right now, just in time. Interventions. So to your point, Joel, when somebody needs help, can we at least nudge them in the right direction to move a blocker? Especially because many of our students are the working adults. You know, our average age is over 35. So you have folks that are working a full day and coming home and then going to school. And you may not have a professor or instructor that's online at 2 o' clock in the morning local time. So if you can nudge somebody in the moment, they need it, fantastic. Now, to Joel's point, also, because it's garbage in, garbage out. There's hallucinations. Everything else we're hearing about, you need human in the loop in a lot of this too. So I've been thinking about more transitionary AI at this point. So how can you have a system with human in the loop where there's a TA or an instructor there and present using the ability of generative AI to be able to go faster and more efficiently and see if they forgot something maybe, and to scale. So there's an efficiency component, but making sure there is a very obvious human presence that's there.

Excellent. Yeah. So it's not just automate everything. You've got some kind of human if people want to speak to humans as well.

Okay, can I give you one example? And if you're a data scientist that's struggling with the storytelling abilities, one of the first things I'm thinking about with generative AI and large language models is ask ChatGPT or something in a controlled environment to rewrite what you want to tell the business in business friendly terms. If you're struggling with storytelling, you can use some of these things to say, hey, this is still a little too technical. ChatGPT or equivalent. Can you give me ideas for other ways that I can say that? That in itself makes a data scientist a superhero when you are overly technical. So I don't think we necessarily have to think about how do you go full blown 100% chatbots, there's no humans or anything, but instead, how can you use this technology to augment what we're doing to help us with our shortcomings.

I've definitely second that use case because I mean, my background's data science. I love writing about data. I have to write a lot of marketing materials as well, which I'm much less good at and just being able to rephrase, get AI to rephrase things so it sounds more marketing. That's, that's a very good use case. Excellent. So I guess Joe, you mentioned like teachers picketing when calculators came out because students can just pass their math exams with by using a calculator. And now it's sort of mainstream. I guess the same thing's happening now because ChatGPT can pass an awful lot of exams. So how has that affected the assessment and certification side of things? Actually. Yeah, Joel, do you want to go?

Sure. So it hasn't had as much of an effect as you might think because I think every university has some form of proctoring of exams one way or the other. So if you're in person in a classroom, there's someone standing there making sure that people are passing the exams on their own and not with outside help. And when, when people take exams online, we have online proctoring systems to do the same to make sure that, that people are passing the exams on their own without outside help. And so either way, either scenario it's really tough. And for most cases to pull up CHAT GPT and ask it what the answer to this question is without being detected. And so I think for other assessments, for take homeworks and projects and things like that, it's a little bit more difficult. At Georgia Tech, some folks have built some pretty good AI detectors to deal with that, but for major exams and things, I think there's not been as much effect so far.

Okay, Joe, have you seen a similar thing? Is it just a lot of proctoring to make sure people don't use ChatGPT when they shouldn't?

Yes, absolutely. So everything we do here is virtual. So there is a lot of virtual proctoring services that are monitoring for other things that may be running within an exam. And you have that virtual environment that's very controlled. But to Joel's point, it hasn't had as large of effect. And I think that's because there's also integrity, which is great. You know, most of our students have great integrity. They actually want to learn the material, they want to be able to get into a role and hit the ground running and they realize that cheating is not going to necessarily be in their favor. Now bearing all that in mind, one of the things I'm thinking about with my leadership team is if folks are actually going to be using this like a calculator in their day to day life, how do you rethink assessments to where maybe you're asking, okay, well what prompts did you go through? Walk me through, kind of the way my 5 year old has to talk through how they're going through math and so forth and actually explain what prompts do you use? What does the business process look like that you're trying to influence? Why did you end up using this particular coding language versus a different one? Why did you use this package or library versus a different one and starting to get into more the effective use of the tools that will be there versus saying, you know, almost living in this very time environment that you're not going to have these things available to you.

Absolutely. Just feel like it's going to be very widespread in the near future and more accepted if, I mean, if it's not even already calculator away from us.

Today, like how realistic is that?

Right.

You know, I liked it. I like to think that I can still do basic math, but I think at this point I've actually forgotten how to do basic math because of calculators. But you know, if these things are going to be prevalent in there. I think the challenge to all of us within an academic setting is, well, what is it that we really want to assess when we're thinking about getting folks ready for the job market?

Absolutely. And so, well, you mentioned the job market, so maybe let's talk about jobs. We've talked about a few different roles people can go for, but what roles are you seeing your students applying for after graduating? What are the most common data roles? Yeah, Joe, do you want to go again?

Sure. So I think it depends on whether it's the undergrad or the master. So in the undergrad space, we try to guide folks towards data analyst and business analyst roles to get a foot in the door. And a lot of that is because when you're coming out of that undergraduate mindset, you're still trying to find in a lot of ways what your passionate is and what you want to do and start building your portfolio, your work portfolio, that's going to go with it. So if you can land that kind of an analyst role, meaning again, those three levers, they're kind of in the middle place. You have a little bit of data, a little bit of math, a little bit of business, you will decide in your career journey from there where you're going to go. Undergrads that try to go directly into the more advanced data science, machine learning, operations roles, devop roles, and et cetera are usually met with disappointment unless they have prior work experience. So there's a little bit of level setting expectation in the masters. It really depends on the track that you're pursuing. But with that in mind, I also want to talk about the career switchers for a second. So again, if you're coming from being an English teacher or a bus driver or something, and you come into analytics, we welcome you with open arms. But there's also an understanding that if you're getting a master's degree and have no relevant work experience, you may not be competitive for a senior data scientist role yet. I don't want to say that you have to do your time, but you need to actually experience it in a business setting to be able to show, okay, this is what's going to happen. I need to build that trust and that repertoire and you will advance into those more senior roles faster because of that incremental training that you had. But there's a level setting expectation of how do you break into these teams and ultimately get to those really high ROI jobs.

Okay, so you're saying basically you get the, the qualification that's going to give you maybe a Foot in the door for an entry level position. In order to stand out, you need to sort of demonstrate that you have on the job experience or some kind of real world skills or portfolio of some kind.

Yeah, yeah. Right.

Okay. All right, yeah, let's. Let's talk about portfolios then. So, Joel, maybe do you want to talk about this? What, what can help you demonstrate that you have those real world skills?

Yeah, so experience obviously is one if you've got the work experience in the field already, but otherwise a portfolio really is important. You can get a portfolio in different ways in most of our programs, including our sort of flagship master's program, we've designed some of the courses so that you get portfolios out of that, where you can point people to GitHub sites where, where you've done these really significant pieces of work and they can actually see your analysis and see what you've done, not just have you write a line on a resume. And I think those are really important. I think getting some sort of experience even before you graduate, whether it's an internship or a practicum course or whatever it happens to be, again, is really important. Just being able to talk to people about things you've done instead of things that you have the potential to do is really a differentiator when you're looking for some of these jobs.

Okay, that is excellent. I'd be slightly remiss not to point out you don't have to use GitHub if you're a data camp user. You do have a personal portfolio within your profile, so please do make use of that. We've got a webinar coming up on February 13th to talk you, talk you through doing that. Just to expand on this, Joe, do you want to tell me, like, what sort of things might go into a data portfolio?

I'm going to sound a little bit like a broken record, but first and foremost is a story. I can't tell you how many times I had folks applying for, when I was at Dell and emc, applying for data scientist roles where their portfolio was a mess. Like, there was no rhyme or reason, there was no explanation. It was just a bunch of code, code that was dumped. So the first thing I want to recommend is remember that this is a product you are creating to sell yourself for that position. It is a thing that represents you and your brand. So if you want to be able to say in the interview that you have great storytelling abilities and can work with the business and you make things transparent and then you point somebody to a hiring manager To a portfolio that is just a code dump, you're not going to get that job. Right. So the things that you want to show in there is your ability to tell the story. You want to show breadth that's relevant to that particular industry. So earlier we were talking about finance, for example, when I was in supply chain optimization was really the big selling point. If you didn't have experience with different types of nonlinear optimization, you know, that was going to be a hard barrier to get through. When I was in sales, it was more of the things that you were going to see in the marketing domain. So market basket analysis, clickstream, trying to find more with association rules and recommender systems. So you want to cater your portfolio to what you're applying for. Right. It's part of that story and getting people excited that you're going to be able to hit the ground running on their team and making sure that you can use the relevant lingo for that industry for sure, too. If you're switching between jobs and just applying broadly across industries and you're not custom tailoring to what you're trying to do, you will look cookie cutter in that regard and you won't stand out from the other applicants. So again, your challenge here is how do I stand out and to make sure that my portfolio reflects whatever I'm saying in the interview.

Excellent. I like the idea of including some of the relevant jargon and lingo just to make sure that it's clear that you understand what's going on in that industry. All right, so we've got some questions from the audience, but before we go to those, just have just one last question. Do you have any final advice for people who are wanting to learn more about data? Joel?

Yeah. So, I mean, there are a lot of options for where you can learn about data. What level, what type of program it is, what type of certificate it is, or certification or degree. There's so many good options out there. Really do the research and find the one that fits your learning needs and your schedule and your budget and go from there. There's lots of good options.

Excellent. Yeah. So not everyone has three years to take a degree, so maybe you need some kind of short program. I like that. And Joe.

Well, first off, welcome to everybody who wants to join the data analytics community. We'd love to have you. Regardless of what your background is, it is a true melting potential. The second thing is have a plan and be ready for that plan to change. Analytics, as an umbrella term, is becoming as broad as saying science or math. And in A lot of ways people use AI the same way. It's so many things underneath that in so many roles. So find a place to start. Be okay with that path changing, but try to have your end goal in mind. If you think you're going to want to spend more time in programming, make sure your path is aligning to that. If you think you're going to spend more time working with the business, make sure your path aligns to that and be ready to realize that you may change them midway through. Never thought I was going to be interacting with the business when I first started. I thought I was going to be at my desk programming the world away. But now I actually prefer being with business people and problem solving and doing whiteboard sessions and then figuring out how the data scientists can help to build something to support that. I wasn't expecting that. So be okay with change.

Absolutely. It's very difficult to predict a career in data, I think. All right, thank you both for all your answers there. So for anyone in the audience who still has a question for Joe or Joel, please do ask them in the chat. We've got a few questions lined up. Before we get to that, I just want to say we've got some upcoming webinars that you might want to attend. So tomorrow we got a session on how to win a data camp competition. So data camp competitions let you do competitive data science and it gives you something to show off in your portfolio. So we talked a bit about portfolios before. Showing that you've won a data camp competition is pretty amazing things. We've got four past winners of competitions coming along to tell you how they, how they did it. On Tuesday, we've got a session on building multimodal search with weaviate. So this is some generative AI stuff. This is how you do search with images. Very, very hot topic at the moment. And then next Wednesday, we've got a session on creating effective graphs. So if you want to improve your data visualization skills, please do come along to that. All right, so onto audience questions now. So first question comes from Kavita. So Kavita says, what are the projects you recommend based on data science? So we talked a little bit about, like, having some kind of projects to show things off. Are there any sort of like, good, sort of broad data science things that are going to help you get a job? I don't know. Who wants to go first?

Yeah, I can think about in terms of when people were applying to roles that teams I used to work for, I in particular was looking for somebody that could take something soup to nuts. So end to end. So I didn't really care what the example was, but I wanted to be able to see that there was a problem you were starting with. You showed how you were thinking about it and how you landed at the approach you were taking and then be able to show what the results are and interpret those in a way that makes sense. Not just saying, okay, I need to put all the flashy algorithms onto a page, but instead saying, I'm going to do this with purpose. And to be quite honest, when I was hiring for the data science team at Dell, you could have used the decision tree in that example and I would have been excited if you took it soup to nuts and being able to tell that story. I want to see your curiosity. I want to see how you look at it from a 100 foot view and a 10,000 foot view in a cohesive way that shows me that you're going to add value to the team. So I also used to see portfolios occasionally that again had everything under the sun. And the first thing that told me is that that person is a master of nothing unless they have a 10 to 15 year resume. Pick a couple of things that are very surgical, potentially aligned to an industry if you can find a relevant data set or example, but make it a complete picture.

I like that. Yeah, because I mean really, certainly my experience of hiring has been looking at portfolios for maybe two minutes and going, this is just a sanity check. Does this person have some skills or have it's been lying on their resume. So I like the idea of just having something surgical that just demonstrates things that are related to a job advert. Joel, do you have anything to add on that? Have you got any advice on creating projects?

No, I think I was going to say the same thing Joe started with. Really make it a really strong end to end piece of work that you put on there. Don't just say, hey, I took these columns of data and ran it through this algorithm and got this output that's not going to be so helpful. But really start with the question you're trying to answer. Talk about where the data came from, how you got it, how you handled it and why, and how you chose the analysis method you chose and why and how you're interpreting the output and why and really showcase all your skills, not just the really technical ones.

I like that having some kind of commentary on why you've done stuff rather than just saying, okay, these are the results.

Nice.

Okay.

All right, so next question comes from asea. So Asia asks like, what's your suggestion for someone who's a statistician? Okay. And which to me. Oh, so how basically, I guess, how do you transition from biggest statistician to data analyst? Oh, maybe you can just talk about the difference between the two roles. All right, who's looking like the one to answer? Joe, do you want to go first this time?

Sorry, Joe.

Oh, either. Either one of you speaking.

You know, the first question I have is why are you making the change? Because you will find statisticians within data analytics teams today. So are you making the change because you feel like you're not working in the data enough and you want to go play in a data lake or a data lakehouse or a data fabric layer? Do you find that you want to do more with visualization? Do you find that you want to be interacting with the business more? Do you want to be operationalizing algorithms? Maybe you want to transition into an IT team that's doing model deployment into a customer facing system. Start with why you are making a transition and what you're looking to achieve. Because I always expected there to be a statistician in some way on some of these teams and it's a very valuable role still. So once you can answer that and reflect, then you can start thinking about. Okay, well, if I really want to think about model deployment, well, now you have your pathway. You need to go focus on MLOps workbenches. You need to be thinking about data drift, model drift, the ETLs, and how you're going to be thinking about that aspect.

I'm feeling like I'm learning some important lessons here about being mindful about what you study in order to align it with your career rather than doing things at random. Excellent. Joel, have you got anything to add there? Do you want to talk a bit about difference between statisticians and data analysts?

Yeah, I think it really depends what you feel like you're missing. Do you feel like you have all the skills you need and you just need a piece of paper with a credential on it so that HR will qualify you for the job you want? Or do you feel like maybe if you've had, say, a very traditional statistics background, that maybe you are missing some of those computing skills and coding and data skills on the one side and maybe the visualization and the business skills on the other side and you really do need to broaden and in that case you'd want a program that lets you focus more on those skills because you already have the hardcore statistics skills. I think either one of those would be a fine answer. It just really depends which situation you're in.

Absolutely. Yeah. So it sounds like there's, you got that one lever pulled up on the stats and you need to pull up believers on the, on the business skills as well. Okay. All right, so I think we've got time for two more questions before we wrap up. This comes from woodworking aspirations. That's an interesting name. So I'm meeting some managers tomorrow to discuss training on data analysis with our very good big fun in the R language. So I've got to learn some good communication skills since some of them have minimal math background. Okay. Yeah. So maybe do you have any advice on communicating to people with no technical background? That's like maybe one of the trickier aspects of communication. Who's going first? Joel. Joel, this time you go first.

Yeah, so. So I like to. I mean you're not going to be able to teach them to, to start running complex things in R if they don't have a strong background. So what I'd like to do. A lot of the mess, the methods in data science that there are good R libraries for have a pretty intuitive background behind them. So like in one of my introductory courses, I start with support vector machines. And it's a really complicated sounding name which really means just draw a line between two sets of things and so you can show that and then give some real life examples. Like if you're out in the woods with your kids, as I've been, and we see a mushroom, so we need to see which side of the line is it on? Is it the poisonous side or the non poisonous side? And just give examples that have maybe nothing to do with statistics, maybe nothing to do with data science, but just ask the same sort of question and answer the same sort of question in real life. I think that's, that's really a nice hook to, to get people in who may not have the background otherwise.

I love the idea of using analogies that people can relate to in order to explain things. I also really like the definition of support vector machines. It's just drawing a line between two sets of things. That's brilliant. I may have to reuse that one. Excellent. Joe, do you have any more advice on communicating to people without technical backgrounds?

I used to train folks that were coming into their rotational programs at Dell on how to use R and things like that too. And these were folks that didn't come from analytics programs or math backgrounds. You know, they may have had a math course in their finance degree. And there were two things I always did. One is I made it relevant to their role. Specifically, why do they even care? Why are they asking you to do this? So if you can use a business case that you can back into that's from their domain and, and is sparking the interest in why they want to go into R, great. And then the second thing is go with algorithmic thinking. You want to show them how to fish, you don't want to just give them the fish. So if you can show them how to use Dr. Google effectively, introduce them to stack overflow, make them self sufficient and then at the end you can go through a basic example. And I used to use a decision tree and I did that because the visualizations were very easy to understand and I would do it for something in their particular business domain. So relevancy, make it real for them. Why are they there in the first place? Algorithmic thinking. You know, how do you think through programming in general, especially if it's the first time they're like they're downloading R. They've never done hello world, anything like that. Start simple and then at the end you can show them a more basic example that's fully thought out. Once they know how to fish.

Excellent. Yeah, I love that thing about making it relevant and I suppose, yeah, I guess whenever you're having a meeting with someone at the beginning, they're not going to care about your analysis as much as you do. So you've got to really make them care. All right, so final question from Oscar. This looks like a yes, no question. So hopefully it should be pretty quick. Can projects and homeworks I've worked on for my degree be considered part of my portfolio? Yeah. How do you turn your homework into a portfolio item?

Don't use the Titanic data set. You know, it's so funny. You absolutely can. As long as you make it unique to you and you tell that story. If it's, you know, the same thing that everyone else. Think of it this way, if 25 other people in your course are going to be applying for that same role and they're all going to use that example, why is the hiring manager going to pick you? You have to be able to show something unique, specialized, something you did exceptionally well. Otherwise you run the risk of it being cookie cutter.

Okay, I like that. Make it a little bit unique to you. Joel, do you have anything to add to that?

Yeah, no, I, I agree. It's. If you're in a program or a course, you're going to have a lot of homeworks and projects, and you're not going to show them all in your portfolio. So really pick out the ones where you did something extra to shine compared to what everybody else might have done.

Absolutely. And so maybe getting that story there is going to be the thing that helps you stand out. All right, brilliant. I think we're at time now, so let's wrap it up. Thank you, Joel and Joe, for all your answers here. Really informative stuff. And hopefully this is going to help people get inspired to learn some data skills.

All right.

Thank you both for your time.

Thank you.

Thanks for having us. Excellent.

And thank you, Reese, for moderating. Thank you to everyone who asked a question. Thank you to everyone who showed up today. Hope to see you all again in future webinars.