CDO Interview: Celeste Fralick

CDO Interview Questions:

Q: How did you get started in data management/data analytics?

As a bench microbiologist recently graduated, I wasn’t making enough money to sustain a life I was dreaming of, so I applied blindly to Texas Instruments on my “analytical” skills in chemistry and micro. I was hired as a quality engineer and my first assignment was to implement statistical process control (SPC). It was love at first sight with data and statistics.

Q: How long have you been in data leadership?  

My first stint as a “chief data scientist” was with Intel in 2015 in the Internet of Things Group as well as leading the corporate-wide analytic Center of Excellence, but I have been working with data for 40 years. Why do I stay in it?  Because it’s fun, you’re changing the world, influencing young careers and seeing positive business impact. 

Q: Recent stats show that most CDOs have been appointed in the past several years. How have you adjusted to coping with these new responsibilities? 

As a chief data scientist, I must remind myself to step above the data crunching, which I love. It is important to examine the systemof data, the management of data, how it can be monetized and valued for a company’s bottom line.

Q: What qualifications do you think the role needs in the industry? 

Domain knowledge, math and stats, data science (ML, DL, AI), software, and leadership. It does help to experience different market segments, as one machine learning model specifically used in, say, earthquake physics, can be quite applicable in cybersecurity.

Q: As a follow up, what would you rate the top CDO Skills to be and why? 

Leadership and statistics. Without 1) modeling the way; 2) challenging the process; 3) encouraging the heart; 4) enabling others to act; and 5) inspiring a shared vision, you’re unable to lead and influence those around you (stealing from “The Leadership Challenge” by Kouzes & Posner, one of my favorite books).  The second skill, statistics, is fundamental to understanding data science, and too few data scientists are well-trained in statistics.

Q: Where do you go to get more relevant information for your position? 

From my peers. CDO Magazine has done an excellent job bringing us together in town halls and roundtables to share our challenges and successes (and you didn’t tell or pay me to say that!!).

Q: What are some specific topics you feel are missing or require more focus?

I find that [the topic of] data systems, from soup to nuts, is sorely missing. We are focused on development, governance, quality, and the typical machine learning/AI topics, but, as an industry (of data scientists), we are not focused on an “ISO 9004” look at what we’re doing. This standard is for quality management systems, but we don’t have an equivalent yet for data science. We’ve put aside the documentation, the feedback loops, the connect-the-dots and accountability approaches that make for strong repeatability and reproducibility. It’s not called data art, it’s called data science, and we need to approach it with that in mind.

Q: What are some of the things you are working on today that you see as beneficial to the organization?

First, I appreciate the quickly evolving and bleeding edge of research in all areas of data science and stats, but pure AI is not happening fast enough for me. So, I would like to see much more cognitive learning algorithms applied. Second, as a general term, AI reliability has not been adequately addressed in the data science realm. It mathematically answers the question: How long will the model last in the field in its intended function?  I am confident these two areas, along with data systems, will benefit the organization.

Q: What do you think is your secret sauce to success in the CDO role?

Leading, not managing. I work daily at “enabling others to act” – that leadership pillar is about listening, bringing diversity of ideas and thoughts into the mix, and working as a team.  It’s a pillar I’ve really had to work on throughout my career, so I remain laser focused on it. 

Q:What accomplishment are you most proud of in your current role? 

My all-women team of advanced analytic researchers. It has been great to mentor and enable the growth of their careers. They are like family to me.

Q: What projects are you most excited about?

I am most excited about data systems and the endless possibilities and challenges they bring. We have led the cybersecurity industry in adversarial machine learning and deepfakes, where the latter has been doubling in number about every six months. I am also very passionate about AI reliability, as general reliability and software reliability models really don’t satisfy describing AI reliability mathematically.

Q: What are the toughest challenges you have experienced regarding handling data or digital transformation? 

The unknown unknowns. It’s always been most frustrating to get data from a pipeline that has questionable or unknown manipulations in it, like filtering, missing critical features or less-than-optimal caching. I have worked in about eight different market segments and it’s always those unknown unknowns that bite you and destroy your accuracy. Even with precise project questions up front, there seems to always be something you didn’t know about.

Q: Thinking about a few years from now, what will be the biggest changes we will see regarding data management?

ML Ops will have a larger role to play as we move from a development-centric to a consumption-centric working framework. We’ve all been focused so much on DevOps that ML Ops has had minimal focus. And yet, it is ML Ops monitoring that will help enable AI reliability (as well as building it in during DevOps, of course).  

Q: We all have experienced some dramatic changes and challenges in the past few months due to the COVID-19 pandemic. How did the pandemic impact your company? 

Our company itself didn’t miss a beat. We were able to transition to home offices without much difficulty – it helps that we are software-based. Our global threat intelligence, however, showed a significant uptick in malicious sites surrounding COVID-19 and related topics, such as government programs. 

Q: How is data being used to help meet these challenges? 

Because of the amount of (software) sensors McAfee has globally (there are approximately 70 billion lookups per day, all within 1ms) we’re protecting our customers against adversaries 24/7, including those malicious COVID-19 and related sites. We will never rest because the bad guys won’t.

CELESTE FRALICK BIO

Celeste Fralick, Senior Principal Engineer and Chief Data Scientist for McAfee in the Office of the CTO, is responsible for innovating advanced analytics and analytic processes at McAfee.  She was named one of Forbes’ inaugural “Top 50  Women in Technology (Americas)” in Dec. 2018,  Industry Leaders 2019 “5 Influential Leaders in Cybersecurity”, Insights Success’ “2020’s Most Successful Businesswomen to Watch.”, and, most recently, CDO Magazine’s inaugural “Global Power Data Women” while earning her company the coveted 2020 IEEE CIS Outstanding Organization award.  She has applied machine learning, deep learning, and artificial intelligence to 10 different markets, spanning a 40-year career in quality, reliability, engineering, and data science.  Celeste holds numerous patents and a Ph.D. in Biomedical Engineering from Arizona State University, concentrating in Deep Learning, Design of Experiments, and neuroscience.