How Our Machine Learning Team Builds Tomorrow's Tech

At ByteDance, the development of our intelligent technology not only drives the success of our products, but also opens up the future to endless possibilities. Our global Machine Learning Team brings together talented engineers from around the world who have a wide range of skills and experiences, as well as a shared passion for working with cutting-edge technology.
Introducing Our Engineers

Pujitha AK is a Recommendation Engineer who moved from her home country of India to join ByteDance in Singapore in early 2021. She has a Master's degree in Image Processing and Deep Learning, and previously developed speech recognition technology for a global electronics company.

"I joined ByteDance because of the opportunity to make an impact on users," Pujitha says. "I believed the work would be more fast-paced and on a larger scale than anything I'd done before, and so far that's been entirely true."

Matthew Goodman is a Research Engineer who joined our Intelligent Creation Computer Vision (ICCV) Group in Mountain View in August 2020. Before ByteDance, Matt had experience with platform engineering for an experience management company and earned a Master's in Computer Science from the University of Pennsylvania, where he also worked as a research assistant in the GRASP lab and focused on leveraging deep learning to synchronize motion-based videos.

"One of the first things I came across in my graduate research was TikTok videos (looking at the motion of people), so I was already familiar with the products ByteDance was building," Matt explains. "I also had respect for the publications I was seeing from the team, which were tackling the most exciting and innovative topics in the field."

Given the range of different roles and projects within the Machine Learning Team, engineers have the freedom to take on responsibilities in different areas and learn new skills. Based on their past experiences and interests, both Pujitha and Matt have found a specific area of expertise that they focus on in their day-to-day work.

Pujitha works with the TikTok data team on developing the app's recommendation system. Specifically, she works on strategies to extract videos from a large pool of videos based on users' interests, and helps build the video and user features that make recommendations more personalized. Her work is end-to-end, which means that after developing a new feature, she also drives the initial user testing and eventual implementation. "Every time I run a test, it can involve a million or more users," she says.

One particular project that Pujitha led focused on TikTok's recommendation system in Pakistan and Bangladesh. "Our goal was to retain users for a longer time, so I created and tested several different features in the region. When the most successful feature went live within weeks, that was the first time I realized the huge impact I could make on the in-app experience."

Focusing on TikTok in Southeast Asia, Pujitha dedicates a lot of time to learning all the local factors that influence the recommendation algorithm. In addition to this project, she also works with our Trust & Safety Team to improve the risk software that helps keep our in-app community safe.

Like Pujitha, Matt drives projects that have a direct impact on our products and users. Striking a balance between pure research and engineering, he works to improve the automated machine learning algorithms behind several of our products and build a general deep learning framework that can be rapidly deployed to produce new research. He also does more traditional engineering work by building machine learning infrastructure and other tools that can play an essential role in future products.

"For any project I'm working on, the research I'm doing can directly affect millions of users around the world," he explains. "For example, in the past we released an automated machine learning algorithm that improved our content moderation by 3%, and that made an immediate difference for all our users. If you improve an algorithm by even 0.1%, that represents a massive step forward for the wellbeing and safety of our community."

Work at ByteDance is fast-paced and challenging, which creates countless opportunities for engineers to improve their skills and develop new ones. "We're growing so quickly and we have so many ideas that there's always something new to explore," Matt describes. "In my time here, I've worn many different hats and had the chance to dive into different fields, like video understanding. Even when you join a project that's outside your usual focus, your perspective is valued and it's all about learning from each other."

Autonomy is a defining aspect of our culture, so we encourage and support ByteDancers who want to engage with topics outside of their comfort zone. For her team, Pujitha cites the tech sharing sessions held each week, where they go through and discuss the latest developments in recommendation systems, including the most recent academic papers. She also highlights the ByteDance mentorship program as one of the main reasons she's found success in an entirely new area.

"When I started at ByteDance, I'd never tested features on users or taken them into production before, so I was confused about how to utilize metrics," Pujitha explains. "Fortunately, my mentor was there to check in with me every day and answer my questions, no matter how big or small. In just a few months, I've learned so much about an area that I'd wanted to explore for a long time."

So what does the future of our team and technology look like?

"My managers have been clear about opportunities for me to grow within the company, and senior leadership meets with me each month to discuss how my work contributes to our long term vision," Matt says. "So while we're always pushing each other to do big things and dream big, we still have a grounded sense of direction, which gives me confidence in the products we're aiming to launch in the future."

Pujitha is particularly excited about the potential of the technology behind TikTok. "When I see that each engineer has ownership and can contribute to each part of the algorithm, I'm certain that TikTok will continue to improve rapidly. The speed at which we work, the new ideas constantly being explored and tested...this is what sets us apart!"

By improving technology at the cutting-edge of our industry, our Machine Learning Team is pushing for innovations in technology that will serve as the foundation for our future success. If you're interested in joining Pujitha and Matt on their global team, make sure to browse opportunities on our jobs page and follow ByteDance on LinkedIn to stay up to date with our latest openings!