Meet our Technical Experts: Brittany Bogle from the Data Science and AI Elite Team
In just a short year and a half she’s helped build a demo to assist physicians in India to prioritize patients that are at higher risk for cardiovascular events and detect model. For Nedbank, she was part of two use cases to help their data science team optimize the bank’s ATM maintenance processes and predict fraudulent banking activity. Her first client engagement was building a model to help healthcare providers at Geisinger health identify high-risk sepsis patients.
Learn more about Brittany Bogle in our new series profiling the technical experts helping clients reach their AI and machine learning goals. Her path to data science elite status – drawing from a multidisciplinary approach to practical problem solving – is what makes her a valuable and unique practitioner.
When people think data science, the impulse is to think coding or mathematics. But what role do less recognized qualities like creativity play in solving problems with data?
Brittany Bogle: Having Python skills, I would argue, is the least important out of the list of data scientist requirements. You have to be able to think creatively and think logically. Half my job—and the most exciting part of my job when I’m working with clients—I have an investigator hat on. We must make sure the thing we are trying to predict can somehow be translated into business. That means we need to be careful when thinking about the inputs going into models. We need to make sure we are being conscientious with domain knowledge gathered from subject matter experts and try to find ways to incorporate that into the inputs for the model.
That part is a creative process and really fun. The more data science can move towards creativity—the better.
You mentioned your “investigator hat.” Could you share with us your process to identify places of client need for data science and AI?
Brittany Bogle: The investigative process is twofold: You must level-set to begin with. You never know what the client’s perception is of data science or what this use case is going to be like. It is very important early on to sit and talk about what data science is—what success is.
It’s also important to emphasize what success isn’t. Rob Thomas, GM of IBM data and Watson AI, likes to say, “AI isn’t magic.” Some clients have that conception that at the very beginning, day 1, we are going to be able to say, "we’re going to improve our accuracy by 20%.”
I have no guarantee that’s going to happen. That’s really the first thing we do. We try to have this conversation and make sure what the client wants to get out of the engagement is realistic.
Second, we make sure that from a high level, we understand what the client thinks they want their KPIs to be. That’s the other difficult part: we have to make sure we communicate early and to the right individuals.
You can define the use case, but sometimes executives don’t know what they have or don’t have in terms of data. Our projects might turn into setting up a data pipeline, because we realize the client does not have the data that they need to answer the questions they want answered.
Then begins an iterative process:
- We form a co-team with the data scientists and the client
- We start digging into the data
- We ask a ton of questions
- We play back our results and get feedback early and often
In the end, you ensure what you are trying to model reflects the defined KPI. If not, you have to go back and have another conversation, adjust expectations, change the metrics; it’s a very data driven process.
We don’t want to make up data or assume too much. We need to ensure that the scope of the project is driven by the data we have available.
IBM has a rich, storied history; because of this, clients may carry some preconceived notions. What have been some common misconceptions clients have had about IBM or Watson AI? They don’t all expect us to be using ThinkPads, do they?
Brittany Bogle: That is something they do bring up! I do have a ThinkPad, but most of my IBM teammates have chosen Macs.
[As to Watson] I hear it consistently, people don’t really know what Watson is. They want to do something with Watson. But it’s representative of a whole ecosystem.
Watson is IBM’s brand of AI ecosystem. That means it could be a natural language processing API for work I am doing, while I have clients right now that call it Watson Studio Local, because that’s what they are working on.
To our clients, Watson means whatever IBM Watson product they are working with. With one client who is using Watson Studio Local, they say: “I’ve loaded this data into Watson!”
From our client’s perspective, Watson is often what [they] want it to be in the context of which tool [they] are using. But again, IBM Watson is representative of a whole ecosystem.
What has been your most memorable client engagement?
Brittany Bogle: Nedbank in South Africa for a lot of reasons. It was a wild engagement!
It was a team of data scientists and, like many data science teams, were diverse as far as skillsets and background. Helping that team to develop a use case, work to overcome data sourcing and model building challenges, and bring each team member’s contributions together to answer the business question was so fun. I remember leaving that engagement and feeling like I was leaving my team. We bonded really well even given any cultural and time zone differences.
A few months later, I received a WhatsApp message. It was a screenshot. They had deployed the model and it was evaluating fraud in real time. They said, “we named the model after you!” That was really rewarding!
Is there a specific client mindset or company culture that produces the best results? What’s the x-factor that clients need to be successful with data science and AI?
Brittany Bogle: In my experience, our Data Science Elite Team model works best if we can form a team and have the client understand that it’s not IBM versus the client’s data scientists. It is not two different teams; we are one team.We are going to work together to get you to the end. Our team is not a consulting group that does the work behind the scenes and comes back with a perfect product or model. Our client data science teams need to walk with us through the development process and own the outcomes. This enables them to grow the data science capabilities after we leave.
This [mindset] is the biggest driver for success.
From the client side, I’ve seen this reflected very interestingly. Whenever that understanding is there and works well, their perception of IBM improves. They start viewing us as a partner and they pass that information very organically to executive leadership.
I think we can give them a different view of where IBM is going.
You can follow our clients throughout their journey to AI in our new client gallery, and learn which comapanies were among the first to confidently put AI to work. Modernize all your data estates and infuse AI throughout all your data, no matter where it lives. Get started with the IBM Data and AI portfolio.