Artificial Intelligence and Rewired Human Connections
Today we’re thrilled to welcome Helen Sun, Engineering Leader at Meta. Helen’s career has taken a unique path from education technology to senior leadership positions at some of the world’s most recognizable companies, including Oracle Corporation, Salesforce, and JP Morgan Chase. Today, her focus is on helping engineering teams develop and implement adaptive, sustainable AI and ML models addressing private-centric data at Meta. These solutions oftentimes require innovative ideas and unconventional approaches.
In this interview episode, we talked about the career transition from education to technology leadership, the challenges of building high-performing teams, the pivot from Big Data to good data, and the promise of AI to automate business decisions.
Watch&Listen to the full interview episode here.
Helen, you have taken a fascinating path in your career, from education to technology leadership. Can you give us a sense of the through line from being a professor to where you are today?
Helen, you have taken a fascinating path in your career, from education to technology leadership. Can you give us a sense of the through line from being a professor to where you are today?
I came from Shanghai, China, where I was born and grew up. By the time I was about to pick a major in college, my mother, who was Head of HR at one of the most prestigious engineering firms in Shanghai, told me not to get into engineering, even though I was strong in all the STEM subject areas. She believed a good career for a girl is in education. So I decided to become a professor. Looking back, I want to encourage everyone to do what they’re passionate about, even if you have to object to the closest people.
When I came to the United States to pursue my Master’s and Doctoral degrees, I found that I was a strong problem solver and decided to step into a career in technology. I gravitated toward data, processing, and databases, which led to Big Data and AI. Now I am at Meta, building world-class tools and platforms to support AI scientists and ML engineers to do their jobs efficiently.
You have a long and successful career in various technology companies in leadership positions. I assume you had to hire a lot of people for different roles. What are some critical elements of building high-performing teams?
Building a high-performing team requires being people-centric. The number one pivot is to focus on people and their strengths. As a leader, you need to leverage strengths, call out the gaps, and find ways to address them. Then you empower your teams by setting a clear vision and allowing them flexibility and freedom to innovate. A leader can scale teams and improve performance by growing people.
Having defined strategies and goals, how difficult is it to allow freedom to innovate?
It’s a very intricate balance between top-down versus bottom-up approaches. On the one hand, you need to understand the industry, have the product knowledge and intuition, and then distill that into a clear vision for the team. On the other hand, you need to recognize that you might not have all the details to be prescriptive about what problems to solve, how to solve them, and at what point in time.
In a successful tech company, engineers have the freedom to choose how to solve the problem and when to solve what problem. They suggest all the prioritized projects at planning time while the leader’s job is to establish clarity of the general landscape because there might be overlaps, duplications, or gaps. The balance of the two approaches and the proper tooling helps the teams execute ruthless prioritization and measure their choices. We’re trying to find ways to support innovative ideas and do it very thoughtfully.
You have co-authored several books on Big Data. Can you tell what inspired you to research the area?
At Oracle, I volunteered to co-author a book called “Oracle Big Data Handbook.” As soon as I got accepted to write the book, I was sick to my stomach because I hated writing. I knew it would be a struggle, but I got better at it and co-authored a second book, “Master Competitive Analytics with Oracle Endeca Information Discovery” (Ed. published by O’Reilley’s). That’s how I started my journey in Big Data.
Now there are many different types of big data – fast data, batch-oriented data, streaming data, and real-time data but I fundamentally believe it’s important to recognize good data. We know that the volume of data is there, and we have various technologies to handle that data. The innovation is more about processing Big Data, whether streaming or batch-based.
Another area of your interest is AI. How would you define AI, and what are the most impactful application of AI technologies available today?
My deep interest in AI stemmed from my work at Oracle. There I led the central area of the Big Data go-to-market and solution engineering team. We worked with some of the most innovative customers in the packaged goods space. I got into AI and machine learning, and various statistical models, including classification, and clustering to help the customers solve category management use cases.
I want to encourage everyone to do what they’re passionate about, even if you have to object to the closest people.
When defining AI, we should start with software engineering because people know what software does – a set of code that runs on a device that will automate certain areas of business processes or personal activities to help efficiency in the business setting. AI, on the contrary, learns from the data to generate patterns. AI learns using data, applies data to different learning frameworks, and then generates an outcome so we can utilize the model to make predictions to automate business processes or personal decisions.
At the concept level, you can use AI in any business because every business needs some level of automation. Instead of mimicking human behaviour (what software typically does), AI mimics human judgment. Therefore, there are going to be a lot of impactful applications of AI in any industry. I’m most excited about how AI creates new human connections, shortens our distance, and allows for a more immersive experience. AI changes the space and time dimensions, and if used correctly, we’ll have an incredible set of new experiences in the future.
In the recent interview with Beena Ammanath, Executive Director of Deloitte AI Institute, we discussed how many companies overlook the importance of ethical AI while training their models using biased data. What is your perspective on ethical AI applications?
I fundamentally believe that good data is more important than Big Data. The volume matters because that’s how we acquire knowledge. But if we learn from the wrong sources, our perception of the world will be wrong, and our decisions and predictions will also be wrong.
Instead of mimicking human behaviour (what software typically does), AI mimics human judgment.
Transparency is critical in the ML and AI journey, and now we understand data sources, the lineage from data to feature to training to model, and to inferencing. We also understand the granular level of data in the distribution of source data, the distribution of features and how the training data compares to the production data. Because if the data is not well distributed and you train your model using unbalanced data, then there will be a bias in your model. Therefore, understanding your data and the lineage is the number one priority.
One of the critical things the industry working on is privacy-centric AI infrastructure, feature engineering, and ML infrastructure. That way, product teams, machine learning engineers, and AI research scientists don’t have to reinvent the wheel of identifying the purpose for specific data sources. We build that to the infrastructure level, so they can do their job more efficiently and comply with the privacy requirements.