The TWIML AI Podcast (formerly This Week in Machine Learning & Artificial Intelligence) Productizing ML at Scale at Twitter with Yi Zhuang - TWIML Talk #271
Jun 3, 2019
In this enlightening conversation, Yi Zhuang, a Senior Staff Engineer at Twitter, shares his insights from leading the machine learning core environment on the Cortex team. He discusses the evolution of Twitter's machine learning landscape and the role of the DeepBird framework, which integrates TensorFlow for enhanced model training. Zhuang also highlights the newly formed 'Meta' team focused on bias, fairness, and accountability in algorithms, stressing the importance of ethical AI practices amid technological advancements.
AI Snips
Chapters
Transcript
Episode notes
Robot Dogs Soccer
- Yi Zhuang's first ML experience was programming robot dogs to play soccer in college.
- They used ML to optimize the robots' walking gait, a tedious process that involved measuring running speed and adjusting parameters.
From Robotics to Search
- Yi Zhuang's robotics team's success helped him get into Carnegie Mellon.
- He later interned at Google, working on storage system performance, before joining Twitter's search team.
Cortex Origins
- Twitter's Cortex team originated from multiple acquisitions, including MadBits, Magic Pony, and WetLab.
- Initially focused on deep learning research, Cortex transitioned to an ML platform team.