In BU’s Online Master’s Program, Students Learn How Data Science Makes AI Happen
Azer Bestavros, associate provost for the Faculty of Computing & Data Sciences, explains how the online master’s in data science (OMDS) degree program equips graduates to seize opportunities to use AI to benefit their careers and organizations, with ethics top of mind.
It was in 2010, when The Economist exclaimed about the “Data Deluge” -- two full years before Harvard Business Review famously declared data scientists occupied “the sexiest job of the twenty-first century” -- that Boston University laid the academic foundation to produce future scholars and practitioners who could use data, mathematics, and advanced computing to deliver insights and innovations.
That year was when BU established the Rafik B. Hariri Institute for Computing and Computational Science and Engineering to cultivate “a community of scholars who believe in the transformative potential of computational perspectives.”
Today, says Azer Bestavros, the founding director of the Hariri Institute and now the inaugural associate provost for the Faculty of Computing & Data Sciences (CDS), it’s clear that the use of data science and machine learning technologies is ushering a new era in which artificial intelligence is transforming every field of inquiry and every sector of the economy.
“We think of data science as the profession of all professions,” says Bestavros. “Simply put, Data Science makes AI happen. AI is what you get when you use the methods, tools, and platforms of Data Science to devise solutions that leverage the power of data mining and machine learning.”
The curriculum for the 16 month, four-semester program starts with a “data science bootcamp” and continues with coverage of core Data Science (DS) topics, starting with four modules covering the mathematical foundations of DS, core machine learning methods, programming and data management at scale.
This is followed by four other modules covering ethical considerations for DS and AI, experiment design and causality, advanced machine learning and AI, and applications of AI in the field.
Complementing these topical modules, the program also includes two modules focusing on real-world deployment of AI, namely “AI for Leaders” and a supervised capstone project in which students showcase what they have learned by applying data science, machine learning, or AI to a particular situation or challenge, preferably one from their own experience.
It’s not a secret that technology giants, governments and investors are pouring billions into developing uses for AI. The investments make the outlook for job growth “much higher than average,” according to the US Bureau of Labor Statistics, which forecasts a 35 percent rise in data scientist jobs between 2022 (when there were 168,900 positions) and 2032. In the US, the median salary for data scientists was $108,000 per year.
This projected growth in employment is dramatic even though we are still in the early days of AI adoption. Just 3.8 percent of businesses were using AI to produce goods and services in 2023, a US Census survey found. That adoption rate climbed to 5 percent in June 2024. As more skilled data scientists apply what they learn in programs like BU’s OMDS, that rate is likely to continue climbing.
In this Q&A, Bestavros describes the OMDS program, outlines its curriculum and approach to AI, and discusses the career opportunities it offers.
Describe BU’s approach to data science.
Over the last decade or so, with the exploding demand for education and research in computing, every university had to chart its own strategy about how to scale up its computing and data science footprint. BU’s strategy in that respect is unique. At BU, the emphasis is less on computing as a technology and more about how computational and data-driven approaches are transforming various disciplines and professions. It is less about technology for the sake of technology and more about how to deploy technology for impact. So, how can computing and data science transform marketing? How can they impact public health? How can they help us understand climate change and promote sustainability? How can they impact education, manufacturing, journalism, etc.? In many ways this is the distinction between computer science and data science.
BU is uniquely positioned to take on this approach. BU is a huge university with 30,000 students. In addition to the Faculty of Computing & Data Sciences, we have 17 different colleges and schools covering every conceivable discipline and profession. Back in 2010, the question for BU was: can we use that diversity as an advantage as we set up our strategy for computing and data science? And the answer was yes! At BU, we think about computing and data science as the way to excel in all these disciplines. So no matter what you want to do, you want to start with data science.
With that vision, in 2010, the university started with setting up a cross-disciplinary institute, the Hariri Institute for Computing, focused on computational and data-driven research. Then, fast forward to 2014, the university decided to adopt that same vision for its educational programs, launching a Data Science Initiative (DSI) and initiating campus-wide discussions around a forward-looking vision for data science at BU. It was at that point that the blueprints for the Center for Computing & Data Sciences were laid down, eventually leading to the creation of the Faculty of Computing & Data Sciences, or CDS as we like to call it.
You can think of CDS as a school or college, but in reality it is an academic unit like no other. Its mission is to connect and uplift all the colleges and schools at BU around data science and AI.
So, when it comes to our educational programs, our philosophy is not to train students to be generic data scientists, but rather to train them to be the data scientists who want to and will transform marketing, who want to and will transform sports analytics, e-commerce, banking, biotech, education, etc. – you name it! It is in that sense that we think of Data Science as the foundation of all professions. And, in the age of AI, and as I like to say “Data Science makes AI happen” – it is how you make AI happen.
The curriculum for the online master’s in data science program includes modules for both data science (including mathematical foundations and programming toolkits), as well as training in AI and machine learning, and the ethics of data science. How does this reflect your educational philosophy?
Our program modules speak to that. It's really about taking you from the beginnings of what data science is, the mathematics of it, the programming that goes into it, and so on. But then we show you at the end, that it's all about how all this comes together to make AI happen “in the field.” This is why the program ends with a capstone in which students make AI happen in the context of a project of their own. The capstone starts with the students identifying a problem that they care about, and using data science to understand the problem, and finally based on that knowledge, recommend and take actions, the AI. So the journey you go through in the program is one that is literally showing you how data science makes AI happen.
The first four modules give a robust data science foundation covering programming, statistical inference, machine learning, and data management at scale. The next four modules focus on applying that knowledge for AI, in particular experiment design and causality where the impact of AI is tested, and responsible and ethical data science where you make sure the AI acts ethically for good. And finally, the personalized capstone project where you make AI happen for your own cause.
Talk more about how AI fits into the program. At a time when AI is making a splash everywhere, why not develop a program in AI as opposed to data science?
I am not sure I would make it an either or! And, as I said earlier, Data Science makes AI happen. Yes, there are MS in AI programs that are starting to pop up here and there. But, what does credentialing someone in AI really mean? AI is not a field or a discipline; it is a broad term about systems and applications that exhibit behaviors that we often associate with human intelligence — for example, by learning from past experiences, or being able to recognize patterns, etc. In that context, one needs to be trained in how to develop such solutions using the tools and techniques of data science, and to be credentialed accordingly. To use an analogy, one does not get credentialed as a Formula One engineer but rather as a mechanical engineer. It is in that sense that our MS in DS is designed to train students in making AI happen.
Can you provide examples of AI in the program?
The program itself has a ton of AI on display – and I'll give you some examples. After completing some initial modules on the math of data science, on the programming toolkits for data science, and on data management at scale, we quickly introduce students to the fundamental and advanced aspects of machine learning and AI. We offer two full modules (out of eight) on that topic. This is where students will learn about large language models and how to go about building an AI platform, etc. They already know how to do programming and they understand the math that underlies all of this, so these two modules will get them to the core aspects of machine learning and AI.
The next module is on the ethics of AI, and it is very important. It's not enough to know how to put together an AI solution. Did you think about how people can misuse it? Did you think about liability issues? Did you think about copyright issues? Did you think about privacy issues? Did you think about equity? Could I design a system that somehow is not accessible to people with disabilities? That could happen, and the question is how to raise the sensitivity of the learner about these things. So that's a third module on AI.
Yet another module we have is about how AI is deployed in different sectors. In it, students are going to see AI used for sports analytics, they are going to see AI used in drug manufacturing, etc. When they see how it's used in different sectors, they will see the commonalities. Believe it or not, there are commonalities between sports analytics and drug manufacturing. At the end of the day, it is all data science being applied to the development of AI solutions. The end product may be different, but the tools to get there are the same.
Of course, we do not expect that we're going to be able to talk about AI in every field. That's impossible. But by talking about enough fields, the students, whatever their profession is, wherever they come from in their careers, will develop this added capacity or muscle to use DS or to develop AI. They can imagine for themselves what it means to their own project!
There's one last module that is very important for making AI happen – and it's very unique. Very few programs emphasize “Experiment Design and Causality.” How do you know that you're not just dealing with correlation as opposed to causation, like if I do X, Y will happen, versus it just happens that X and Y are correlated? You don't treat cancer by just saying, “Oh, there's a correlation between cancer and smoking, then maybe if I just cut smoking, then cancer is going to disappear.” In our program, we have a module on experiment design and causality. Why? Because if you're going to use data science anywhere, you have to appreciate and understand how to design your solution in a way that does not conflate causation with correlation, and because you will need to know how to measure and/or build confidence, how to do (or interpret) things like A/B testing, etc.
So while there's a lot of AI in what we offer, we chose on purpose not to call our degree an MS in AI. It's about the data science behind it, it's about the programming, the math, and the fact that in any solution, all of these things are integrated. We don't want to think of it as a bunch of unrelated modules, it's really a journey.
How do you help students take what they learn and apply it to their careers?
We do this throughout the program. In the AI for Leaders module students get the opportunity to hear from industry veterans about the practical applications of the methods we teach as well as their economic, social, and policy impacts. Later in the program, you get to try this for yourself! Say you're an employee of a company and you are enrolled in our program to advance your career. As part of your job, you must have encountered a problem that you may be interested in solving by leveraging data at your disposal. Now that you know all about data science, machine learning, and AI, this is your opportunity to apply it all (or some of it) to tackle that problem of yours. You will do this on your own under the supervision of the teaching team, and preferably in collaboration with a team of other students in the program who are interested in solving similar problems. By the end of that module, you end up with something in your portfolio that you can show as evidence of what you built.
What advice do you have for prospective students with little data science experience?
Because data science is a deep field, candidates for our program should have some data-driven work under their belt. It does not have to be work on “big data '' and it may not work done using fancy or complicated software, but it needs to be data-driven nevertheless. Now, if you haven't done any software development before, it is going to be a little challenging. But, if it is just about being “rusty”, then one of the modules in our program – “Programming Toolkit for Data Science” – will help you get up to speed.
Data science also requires some understanding of key mathematical concepts in probability or linear algebra and so on. You might have done that 20 years ago in high school or college, and you may only remember a little, and that is perfectly fine since we have a level-setting module for these key mathematical concepts. It's only one module, but we'll cover enough so that you can appreciate why things work. You're not going to be a mathematician. But you're going to be an informed user of a technology that relies on statistics and math. And we need to understand some things about that.
I would say that having no experience in math or programming is going to be a heavy lift. That's why we have the “Data Science Bootcamp.” The bootcamp is free of charge and is an excellent way for prospective students to assess if this program is the right one for them. So, going through the bootcamp, if you feel like you're totally lost , then maybe the program is not for you. But if you do well, then we are confident that you're going to succeed. Other programs don't do this. They rely on whatever you submit as your credentials. We look at your credentials, but we also give you the opportunity to gauge your capacity to succeed in the program through the bootcamp.
To summarize, I feel that the best students for this are students who are already dealing with data, but want to take their game one notch up.
How should mid-career students considering this program, including those who continue to work at their jobs, think about managing their commitments?
The working professional has to balance life and work and has to answer the question, can I do it? Our online master’s program in data science can be done in as little as four semesters and as long as eight semesters. You can even take a break in the middle. We designed the program with a particular ordering of the modules, but the order allows the modules to spread. It's like an accordion. You can start with one semester of full load and then relax a little because maybe you want a break or because you were in between jobs and now you got a job and can't do two or two and a half modules per semester. It was very important to us in the design of this program to make it actually accessible to qualified students, and that simply would not have been the case if we forced them to drop everything for a rigid schedule.
That flexibility is quite important. It could also be helpful for some students who may lack the programming background. I would suggest they take the module for programming by itself in one semester. Why? Because then we'll give you resources to support your learning. There's other ways you can build it out. Now, you'll end up working more than somebody who already has the programming experience, but that's part of it. You're not going to pay more, but we'll give you the resources so you can learn this on your own.
This is where advising comes in. So when the students come in, we will see what they bring in terms of experience. This program is designed for students to really come from different backgrounds and different preparation, but it has the ability to slow down the part where you need more background and so that you can take your time, but then when you reach other parts where you can go as fast as you want.
What kind of opportunities do you enable for students to build a peer learning community?
The “AI for Leaders” and “AI in the Field” modules will be open beyond graduation. Let's say you take the capstone class. When you present your project, we'll open the presentation to anybody who actually had ever registered for this program. So you can come back five years later and see what students are doing in the program. They're taking the same modules as you took, only five years later, and you get to see the projects they worked on. You’ll be in the audience. You're not going to earn credits. But you are part of that community. You can come to it anytime, any semester, and go back and take a look at these two modules.
The AI for Leaders module is really about exemplifying how AI is transforming different fields and having people come talk about how AI is deployed in their business. So think of it like a TEDx talk, or panel. The students who are registered in a given semester have the front seats. They will be the ones asking the questions. They will be the ones participating. But students who took that module five years ago can still come as an audience.
The important piece here is that this program is designed to connect you with this community well beyond when you take it. You can always go back and see what has changed.
BU’s Faculty of Computing and Data Sciences already has degree programs on campus for undergraduates and graduate students up to Ph.D., including degrees in data science. Why start an online degree program?
The answer to that question is that it is all about democratizing access to computing, data science, and AI, which is the cornerstone of our mission in CDS. Not every learner can take a year out of their life to join a full-time residential program. We see our various programs catering to different learner populations (domestic vs international, early career vs mid-career, etc.) and we designed these programs to recognize those differences and to ensure success.
Learn more about the Online MS in Data Science offered by BU Faculty of Computing & Data Sciences.
By Michael S. Goldberg, Maureen McCarthy