Novetta’s Machine Learning Center of Excellence (ML COE) is an industry leader in adapting deep learning tools to address complex challenges. From facial recognition to predictive modeling, the real-world impact of machine learning algorithms continues to grow. At Novetta, researchers utilize diverse datasets to drive state-of-the-art research and implement cutting-edge tools. As AI4ALL’s newest partner, Novetta is affirming their commitment to ethical development by empowering today’s diverse AI leaders, providing volunteer mentors to our students, AI/data science seminar leads for our programs, and participating in career workshops. Changemakers — AI4ALL’s alumni — can also experience AI in practice through AI4ALL and Novetta’s joint internship program.
Role models widen our horizons, influencing our perceptions about what is possible. In this Role Models in AI interview, Shauna Revay lends her insight as lead of the ML COE, underscoring the impact of representation and mentorship in her own career trajectory. Mentorship is a foundational tool AI4ALL programs utilize in creating space for historically excluded students to envision themselves in the AI field. As an AI4ALL partner, Novetta is providing mentors, including Shauna, for our inaugural Changemakers in AI Portfolio Project Program, furthering access for our Changemakers, and supporting the portfolio projects that will propel them into the AI field. You can learn more here about the ML COE’s exciting advancements at Novetta.
Read on to learn about Shauna’s work as Novetta’s ML COE lead, what trait she most values in a role model, and her advice for those hesitant about, but interested in, pursuing a career in the AI field.
As told to Nicole Halmi of AI4ALL by Shauna Revay; edited by Camryn Burkins.
AI4ALL: Can you tell me what you do in your current role as Lead of the Machine Learning Center of Excellence at Novetta?
Shauna: The Machine Learning Center of Excellence (ML COE) is the research and development hub for machine learning at Novetta. We play a couple of different roles. Our job is to stay abreast of state-of-the-art developments in the field. Day to day, we take part in knowledge sharing across the company, whether it’s in the form of seminars or contributing to white papers, blog posts, Slack communication, etc. My job as lead of the ML COE is to facilitate, coordinate, or actively participate in all of those things.
The other side of what we do in the ML COE is develop rapid prototypes of actual machine learning solutions — things that are going to help either our internal teams or our customers. This usually involves implementing and building solutions on open source tools, or even custom solutions. I was previously an engineer within the ML COE for about two years, so I was actually hands-on building these sorts of things. Now, as the lead of the ML COE, my responsibility has shifted to more project management, oversight in code reviews, generating a lot of the project ideas and direction that we’re going to take, and contributing writing samples to business development proposal efforts.
Your recent research has focused on multiclass language identification using deep learning. Can you talk a little bit about that work? Does that research relate to your current role at Novetta?
One of the cool things about my job right now is that I get to explore a lot of different areas of machine learning. The whole gamut of machine learning problems is what we tackle on a day-to-day basis. Audio analysis is one of those machine learning problems. I do research in natural language processing and computer vision graph networks. For this audio project, specifically, I was trying to determine what language is being spoken in an audio clip directly using the audio’s waveform and deep neural networks. Since I’ve done that work, there have been a lot of open source updates in the field of audio machine learning. It has been exciting to pay attention to that field and see how it’s growing and to know that I was doing research in an active area.
That’s really cool. Can you talk about some of the applications of the audio analysis that you were describing?
Consider Siri and the other common voice assistants used these days. To work, they first assume that you have a similar language. Maybe Siri thinks you’re speaking in English, and so she’s going to start transcribing based on what she thinks is English. But in a case where you have multiple languages being spoken, the first step is to make sure that the transcription model, the AI, is speaking the same language that the speaker is speaking. I envision my research as tagging on to the front of some sort of transcription model to help it work for multiple languages.
You have a Bachelor’s in Economics and a Master’s and PhD in Mathematics. How did economics lead you to math, and ultimately to machine learning?
I started out as an economics major in undergrad. I realized quickly in those classes that my favorite part of all the homework and assignments was — the math. So, I actually ended up adding on a second undergraduate major in math. I was a double major because of that, and as I pivoted to math, this is when a close mentorship experience with one of my professors led to me applying to obtain a graduate degree in math.
It was during my PhD I started working at the Naval Research Laboratory in the tactical electronic warfare division. In my PhD program, I was still on the pure math side so I wasn’t dealing with machine learning in terms of my PhD research. The Naval Research Laboratory is where I started combining my math background with machine learning to answer some of the signal processing challenges that we were working on. That was when I got exposure to the field. I really dove in there, and afterward, was looking for a machine learning-based role for the next part of my career.
For most of your career, you’ve been involved in fields that have historically struggled to attract and retain women. You mentioned that as you were doing economics, you realized that you enjoyed the math part the most. Were you interested in math at a young age? Do you have any advice for women and girls interested in pursuing careers in fields like the ones that you’ve pursued a career in?
I was always interested in math at a young age. But as a young girl, I never thought that I was going to get a PhD in math — which is interesting, since I was so interested in it. I don’t think I had any female math teachers in high school, or even in middle school.
It really wasn’t until I had my first female math professor, in my undergraduate studies, that I remember thinking, “wow, she’s so good at this.” And it dawned on me that I had not had a female math professor before.
I actually ended up working closely with her throughout my undergraduate degree. She was really a role model for me, and was one of the big reasons that I ended up pursuing higher education and mathematics.
With that in mind, my advice is to seek out mentorship — whether you’re in school or in the workforce. Some places are going to have formal mentorship programs, but even just fostering relationships with professors or colleagues who you think you can learn from is a great way to get guidance into career paths that you might not have envisioned for yourself from the outside.
Another piece of advice I have, and one of the unique things about this field, is that you don’t need a degree in machine learning in order to start working in, or contribute to, the field. In my office, those who work on machine learning come from — computer science, math, physics, engineering — all different backgrounds. Don’t be afraid to pivot if you find something interesting. There are a variety of paths that you can take to break into the field. It’s exciting.
Did you have role models growing up? Do you have any now?
Growing up, I always had an admiration for people who were really passionate about what they were doing. Both of my parents had that, and I would see coaches and teachers giving off that same vibe. I always felt like I wanted to find something that sparked that same level of joy for myself.
Now, there are so many role models that I have — from a small level, like people I work with, up to an international stage. The most exciting part is, compared to even when I was younger, there are more examples of strong women in different fields and positions than ever before. I think that’s really inspiring for the next generation.
What has been the proudest or most exciting moment in your work so far?
I think it has to be the day I defended my PhD thesis. It was my last semester of school and I was fighting morning sickness in my second trimester of pregnancy. My PhD is something I had been working toward for years, and it was a goal that I hadn’t initially envisioned for myself. I am proud of myself for achieving that.
What, in your view, are some of the things people should be doing now to create a positive future for AI?
It’s been such a changing and advancing field over the past couple of years, that I think we researchers have been busy trying to create the fastest, highest performing, and biggest models. What we need to do now is take a breath, and think about the underlying ethics of what we’re creating.
About Shauna
Shauna Revay, PhD currently leads the Machine Learning Center of Excellence (ML COE) at Novetta. In this role, Shauna leads a team of ML engineers who develop rapid prototypes to solve some of Novetta’s customers’ most challenging problems using machine learning. Shauna actively works to stay abreast of breakthroughs and emerging technology in the field of AI in order to continue to shape and manage innovative projects that will be impactful to our customers in the government contracting space.
The ML COE also aims to foster internal knowledge sharing throughout the company so Shauna works with her team to run seminar series, and write newsletters, white papers, and blog posts. Shauna also takes part in many business development and proposal efforts at the company in order to help secure new work where Novetta can apply its ML knowledge.
Before leading the ML COE, Shauna was an ML engineer at Novetta for 2 years, where she actively developed ML prototypes and presented new capabilities at the company. Before working at Novetta, she was a mathematician at the Naval Research Laboratory (NRL) in the tactical electronic warfare division. During her time at NRL, she was exposed to various deep learning techniques that her team utilized for signal processing applications. This sparked her interest in AI and she sought out a position at Novetta.
Shauna holds a PhD in mathematics where she studied non-harmonic Fourier analysis. Shauna enjoys using her math background and applying it to different domains such as electronic warfare during her time at NRL, and to machine learning in her current position.