Bridging the gap between research and production for machine learning

About this event

The 2020 Dean's Research Seminar Series kicks off with guest speaker Chip Huyen who is working with a US based start-up that focuses on the machine learning production pipeline.

Humanity has come a long way from the days we were looking up to the stars to “predict” the future to predicting human behavior with machine learning. The global AI software market is expected to grow from $9.5 billion in 2018 to an unbelievable $118.6 billion in 2025. However, embracing AI and machine learning is just the first step for modern businesses. They must morph into AI-driven enterprises if they intend to remain sustainable in the increasingly competitive landscape. IT professionals are the primary force to enable such enterprises by deploying and managing machine learning models. But, productionizing machine learning comes with its own set of challenges." (1)

In this presentation, Chip will explore the key differences between the research and production environments:

• data
• goals
• compute requirements
• evaluation metrics

These differences - together with the lack of communication between researchers and product engineers, the misalignment of business performance with technical performance, an undeveloped infrastructure, and the lack of a feedback loop - make it challenging for organizations to productionize promising research.

Chip breaks down the different phases of a machine learning production cycle, the infrastructure currently available for the process, and the industry best practices. She also addresses the need for interpretability, safety, and privacy, and the trade-off between efficiency and fairness when decisions affecting real people’s lives are automated using big data.

ABOUT CHIP

Chip Huyen works to bring the best engineering practices to machine learning research and production. Her experiences include Netflix, NVIDIA, and Primer. Previously, she created and taught the Stanford course TensorFlow for Deep Learning Research. She’s currently with a startup that focuses on the machine learning production pipeline.

When not teaching machines or helping companies set up their in-house machine learning teams, she writes. She’s the author of four bestselling Vietnamese books, and her writings have been published in leading newspapers in the USA, France, and Vietnam. She’s currently working on a book on Machine Learning Interviews.

Following the seminar there will be the opportunity for networking over light refreshments.

(1) https://endpointlearning.com/blog/2019/6/25/4-challenges-to-productionizing-machine-learning