Meet Erin LeDell, the Chief Machine Learning Scientist at H2O.ai, an open source software platform that provides products that use AI to do AI: automating AI systems to make it easier for everyone to have access to AI based tools. She is also the founder of Women in Machine Learning & Data Science (WiMLDS), DataScientific Inc., and the co-founder of R-Ladies Global.
Erin is passionate and active in supporting and promoting women and gender minorities in machine learning, computer science, and data science. She believes that creating good support networks through meetups like WiMLDS are important for minority groups in terms of opportunity and diversity in the field of AI.
Learn more about her journey as a mathematician, her passion for creating strong communities in the field, and her thoughts and ideas on what we should be doing to create a more positive future in AI.
We interviewed Erin as part of AI4ALL’s Role Models in AI series, where we feature the perspectives of people working in AI in a variety of ways. Check back here for new interviews.
As told to Eunice Poon of AI4ALL by Erin LeDell
EP: As the Chief Machine Learning Scientist at H2O.ai, what does a typical week look like for you and what kind of responsibilities do you have?
The main project I work on at H2O.ai is the H2O AutoML (Automatic Machine Learning) algorithm and tool. For the first few years that I worked at H2O, we focused on developing all the algorithms that we have in the H2O library. Now we have another layer on top of that which automates a lot of the data science process. This is what we call “H2O AutoML” and this is the team that I lead.
If there are questions about machine learning (ML) algorithms internally within the company or externally with a customer or person from the open source community, sometimes those questions come to me and I help to resolve them. But mostly my job is to think about and experiment with what we are going to build next — what kind of algorithms or functionality we should add, what the software interface should look like, and what makes sense from a holistic algorithmic point of view for H2O.
EP: You have an extensive background in mathematics. What got you interested in math, and how did you learn about machine learning and transition into the field?
From a young age, I remember really liking math. It was the only thing I wanted to do in school. I think math makes me happy because there is typically a right answer. It’s also very straightforward and interesting to me, so I decided to pursue a math major in undergrad. After undergrad, I started pursuing a PhD in math, but then I decided that I liked coding too much and I wanted to be on computers more, so I dropped out of my PhD program, took a Masters degree, and became a software developer. This period of life for me was where I went off and steered away from math-related topics for a while.
Learning about machine learning was really exciting for me, because that’s when everything clicked. In isolation, math and CS were interesting, but not exactly what I was looking for. With ML it felt like a really good fusing of the two: my math and computer skills.
After I discovered ML, I wanted to study it more in depth so I looked into PhD programs and learned that you could either do a PhD in CS or Statistics. Coming from a mathematics background, I thought statistics made more sense to me, so I pursued a PhD in that.
EP: You founded Women in Machine Learning & Data Science (WiMLDS). Can you tell us what prompted you to start the organization and how meetups like these can help to inspire and shape the industry?
Organizations like WiMLDS are exposing more women to the field and creating a support network for women so they can grow their career, provide mentorship and networking opportunities, help women learn about different topics, and find jobs in the field.
WiMLDS came out of me attending WiML Workshop, which is co-located with NeurIPS every year. I went to this workshop two years in a row, and after the first one I thought that it would be great if something like this existed locally and in more of a meetup setting.
Meetups were a huge part of my life early on in my ML career. Before I went back to school to obtain my PhD, I co-organized, one of the first ML meetups in the world back in 2008 at a hackerspace called Noisebridge in San Francisco. I spent a lot of time there and got great hands-on experience with ML techniques, and I thought this type of setting would be great for creating a learning environment and support network for women.
After my first WiML Workshop, I ended up waiting a whole year to see if someone would start a local chapter of some sort that was associated to WiML, but nobody did, so I decided to start the Bay Area WiMLDS group. At the time I knew one of the organizers of the WiML Workshop and we talked about the idea of having “chapters” of WiML, but I ended up expanding the scope a bit to include data science topics and focusing more in applied/industry topics vs pure research.
When I started WiMLDS, I thought it would be important to include other things in data science that aren’t necessarily ML, like data engineering for example. I also wanted to make sure that it was more inclusive, so that’s why I tacked on the extra DS to the end of the name, which stands for data science.
EP: Who were your role models growing up and do you have any role models now?
I grew up in the pre/early internet time, so I didn’t really know or learn about any historical women in science, and I didn’t learn about women in school or the history of science in general. Now, we are lucky enough to have lots of information on the internet about inspiring women in science!
The role model I had growing up was my mom. She is an epidemiologist and I thought that she had an interesting job. Her biggest advice to me growing up was to choose a job that I enjoy because I was going to spend most of my time in life working. “If you love what you do, you will never work a day in your life” is a quote she would tell me often.
In terms of role models — Bin Yu, Fei-Fei Li, Daphne Koller, Tamara Broderick, Rachel Thomas, Anima Anandkumar, Kate Crawford, Timnit Gebru and Joy Buolamwini come to mind. I also think Hanna Wallach and Jenn Wortmann, the co-founders of WiML Workshop, are big role models for me — I have a huge respect for their vision and execution of WiML Workshop and the work they continue to do around inclusion and diversity. But honestly there are so many important women that are doing all sorts of amazing work that I am inspired by, so I have a very long list.
EP: What are the types of things people should be doing to create a more positive future in AI?
I guess it depends on which people we are talking about. For companies, I think what they should do is make sure they are hiring a diverse workforce. This is important because if you don’t actively prioritize diversity, it won’t happen. Promoting women into leadership positions is also important.
Once you have the right makeup of the people in companies that produce AI, then we’re going to get better more fair AI systems. We’re also going to get a diversity of viewpoints and opinions in terms of how they’re applied, what ethical considerations are put in, and what checks and balances are created. This is particularly important in protecting minority groups.
If we are using a tool on a large group of people, generally we have to think about how these tools are affecting minority groups because the the scientists might not treat all groups equally when training the models or the data might not contain enough information about all groups.
So I think it goes back to who’s designing the tools and who’s working at the companies making the tools. If you get that part right, then there’s a better chance that things are going to go well.
I would say that it’s important for women and minorities to be the ones creating the systems in AI and not just applying existing systems to different use cases. It’s important to have diverse people creating models and software because once the software is created, then you’re limited in terms of what it can do and how it can be used. So the more diverse minds we have from the start the better.
I also think that bigger conversations have to happen on a government level around AI ethics and regulation. As a society, we need to form rules of how we build and implement the technology because it can have such a huge impact.
EP: What advice do you have for young people who are interested in AI who might just be starting their career or academic journeys?
I would say particularly if you are a young person in a minority group, it would be important to connect with people within your group and other minority groups to create some sort of support network. I think career advancement is more difficult for people in these minority groups — if you feel isolated, then you’re more likely to exit the field or be held back by discrimination, so it’s important to have a good network of support to help you when things get tough.
For young people, AI4ALL is a great group to get involved in. If you’re more at the college level or career stage in your life, then WiMLDS is a key group to get involved in. There’s also Black in AI, LatinX in AI, and many other groups that serve different minority groups. These groups are also important in terms of mentorship and meeting people that can help guide your career.
Erin LeDell is the Chief Machine Learning Scientist at H2O.ai. She has a Ph.D. in Biostatistics with a Designated Emphasis in Computational Science and Engineering from the University of California, Berkeley. Her research focuses on automatic machine learning, ensemble machine learning, and statistical computing. She also holds a B.S. and M.A. in Mathematics.
Before joining H2O.ai, she was the Principal Data Scientist at Wise.io and Marvin Mobile Security, and the founder of DataScientific Inc. She is also the founder of the Women in Machine Learning and Data Science (WiMLDS) organization (wimlds.org) and co-founder of R-Ladies Global (rladies.org).