American Family Insurance has a strong relationship with data science and AI research. We've collaborated with partners at various universities and presented many papers at conferences on topics from computer vision to natural language processing. Recently, we've announced major support of the new Data Science Institute at the University of Wisconsin-Madison. As a research data scientist, I am in a unique position mediating between business sponsors, product owners, partners at universities and cutting-edge research. For this talk, I will offer insight into the research process at American Family and how we integrate new research into products. After introducing the general process, I will talk about some recent work involving knowledge graph-driven workflows for entity refinement for chatbots. Finally, I will speak to some general lessons and best practices (borrowed from software engineering) for bringing ideas from research into production at enterprise scale.
Devin Conathan joined the machine learning research team at American Family Insurance as an intern in the summer of 2016 and came on full-time the following year. He has undergraduate degrees in mathematics and philosophy from Cornell University and a masters in electrical engineering with a focus on optimization and active learning research from the University of Wisconsin-Madison. For work he enjoys developing full-stack solutions that use state-of-the-art machine learning algorithms for industry-quality applications and research. His recent work includes implementing active learning for text annotation, CNNs for chatbot intent-classifiers, and building out an AI-driven knowledge graph platform for powering knowledge-rich applications at American Family. In his free time, he enjoys riding his bike, playing music, and reading sci-fi novels.
Sponsors: I would like to thank American Family for the food at the meetup and Cloudera for an after meetup round of drinks.