Using LLMs for more than chat.

I noticed that good machine learning models did not lead to great products. I started the human centered AI research team and worked with the design studio to understand how machine learning impacts user experience, and based on that understanding, created machine-learning-based products and experiences that people love. We shared what we learned at WWDC and released new human-interface guidelines for machine learning. I eventually moved into the Design Studio and designed the vision for generative computing across Apple products.

I wanted to make it easier for teams to work with data and machine learning, so I started and led the Colaboratory team as part of Google Research, now Google AI. I also got to watch the rise of Google as an AI-first company, help teach the first large scale machine learning and data science courses at Google, and teach one of the first data science courses at Columbia University.

Data. Science. Data Science.

I noticed that machine learning tools were hard to use and often used improperly. I studied machine learning engineers, understood their workflows, and built tools to make their jobs easier. I wrote a thesis that I’m proud of and whose underlying findings are still relevant a decade later.

I wanted to learn more about machine learning, so I started a machine learning focused masters at Stanford. I happened to be in the room when Stanford decided to build a car for DARPA Grand Challenge and got to write the first paper from Stanford on self driving cars. I also moonlit in Jeremy Bailenson's virtual reality lab studying how people learn physical skills in VR settings.

A series of disconnected, seemingly random events and interactions led me to study at Carnegie Mellon, take classes in computer science, machine learning, and design, work with a research team that built speech systems on wearable computers with head mounted displays, and apply to graduate school. These experiences set the groundwork for everything I’ve done since.