Embracing Growth

After a growth-packed 4 years, one of our Machine Learning Scientists, Johnna Sundberg, is leaving Pendulum. She’ll be embarking on an exciting new adventure – a PhD program at one of the world’s top Artificial Intelligence schools, Carnegie Mellon University. 

While this is of course sad news for us, we could not be happier and proud of her. Because not many know, but Johnna never started out as a traditional Machine Learning Scientist. Within a short span of time, she went from an Economics degree and some Data Science experience to key contributor on our most advanced AI team. 

So we sat down with Johnna for one last time, and instead of talking projects, products and business, reflected on her journey to understand what has been fuelling that relentless drive for growth and her career at Pendulum. 

‍

Let’s start with the most obvious question that aspiring AI Scientists have looking at your career from the outside. How did you, starting out with an Economics degree, ended up in the most technical team at a pioneering AI company? 

I think my career just naturally evolved. My undergraduate degree was in Economics. Already then, I took computer science and programming courses because I knew I was interested in working with data in Monitoring and Evaluation (M&E). At my first job, I quickly took on more M&E projects and eventually joined the team once I was certain that I wanted more technical work. From there, I became interested in other ways to use data, especially more proactive ways, and began studying Data Science and looking for appropriate places to apply it to my work. My initial application of Data Science at my previous position eventually led me to Pendulum, where I became excited about the opportunity to dive even deeper into Data Science while still working on socially beneficial applications.

‍

Ok, but Pendulum is not really a Data Science company, right? Our AI team is led by Suvrit who works at MIT and has been cited some 15,000 times if I remember correctly. Plus, the rest of our machine learning team has technical PhDs. So how did you go from Data Science to ML/AI? Or would you say Data Science and ML/AI is the same? 

Good question. I’d say Data Science and ML/AI are interconnected but have distinct focuses. Data Science involves much more manual extraction of data and insights. You are responsible for cleaning the data, integrating and visualizing it, and finally using the right algorithms for predictions. ML/AI is the natural evolution of this process where you are more involved with the development of the models that allow the computer itself to learn from and make predictions or decisions on the data, without having to explicitly tell it what to do. 

And yes, I gradually evolved from Data Science into the ML field. 

When I started at Pendulum, I first focused on very applied work: forecasting vaccines at health clinics, predicting patients who will not show up for care, and identifying shared characteristics among populations that make them more or less likely to show up for care. However, our applied work has many challenges where traditional Data Science methods can fail: we often have limited data, incorrect and even missing data. 

Dealing with these issues is very time-consuming, so I began working more closely with Suvrit to try to use machine learning to automatically build features and work with messy data. This gradually evolved into a process of improving machine learning models to work with tabular data, which is the bulk of our data here at Pendulum. 

‍

Understood, thank you for clarifying! It is a never-ending learning curve for us non-AI people here at Pendulum. What would be interesting to know is what were your most favorite projects you’ve worked on during your 4 years? 

First of all, you are not alone. I think this field is innovating at such a rapid pace that it’s a never-ending catching up for all of us. Which is also what made my time here so exciting.

When it comes to what I loved the most about my work at Pendulum, I’d say it was the projects where I could have a positive and direct impact on people’s lives. I have really enjoyed working on Predict/People as applied to the HIV/AIDS sector. Increasing treatment adherence is incredibly important, so I am happy my work has been able to contribute to this goal.

I’ve also enjoyed researching ways to improve machine learning on tabular data - most data across industries is tabular. Discovering ways to improve performance on tabular data is very impactful for Pendulum and beyond, for the broader AI field. 

‍

Yes, agreed. A lack of meaning is usually not something that I’ve heard our colleagues complain about. Ok and now to my final question, promise. After the breakthroughs in LLMs and ChatGPT, there are many new promising AI companies. So unfortunately for us and fortunately for you, a lot of competition for ML experts and soon-to-be-PhDs, like yourself. Could you reflect a bit on what you think we are doing different here at Pendulum, what should we keep doing and what should we change? 

First, the people! Everyone I have worked with is incredibly passionate and smart. People here are just pleasant to work with. So keep hiring the same kind of people!

Second, is our approach to ML. Pendulum is one of the early companies to have pioneered the use of data-centric AI outside of academia. We often work on ML problems even with limited, messy, and small datasets, which others may consider unsuitable for ML. Creating and improving ways to work with this data is crucial, because most of the world’s data is unstructured, messy, incomplete, etc.

And lastly, what could you guys improve? I think keeping up the effort to ensure the ML team has even more time to work on blue-sky ideas that will lead to advances of the overall field. 

‍

Ok Johnna, as promised - you are released from the interview chair! I think I am speaking for all of us, by saying that we’ve really appreciated working with you and seeing your impressive growth here at Pendulum. This is a bittersweet moment for us. But as author Leo Buscaglia said: “Change is the end result of all true learning.” And it seems like you’ve had a true learning experience here with us, so we couldn't be happier for your upcoming change. We wish you all the best. And maybe, if we are lucky enough, we’ll see you here at Pendulum again!