In this talk, we'll discuss the deep learning–based natural language processing (NLP) technology used by Eno, Capital One's first-of-its-kind intelligent assistant. With Eno, regardless of channel, customers are able to get their questions answered digitally and quickly. We'll discuss the complexities of complex dialogue, why we built our own solution, and present our neural network architecture for NLP. First, we trained word- and character-level word embeddings on web chat conversations. After unsupervised learning for language, we then applied supervised learning to map intents. The result is a model that's robust to common misspellings, abbreviations, and new words. Our model is trained, deployed, and continuously becoming more intelligent over time.
Zach Kulis leads Conversational AI algorithms and software engineering at Capital One. In2017, he led the development of the Natural Language Processing (NLP) stack for Eno, CapitalOne’s multi-channel Conversational AI.
Today, he’s involved in the development of Eno’s nextgeneration AI capabilities. His work encompasses natural language understanding (NLU),dialogue management (DM), natural language generation (NLG), and ML-assisted data labeling.Before joining Capital One, Zach developed ML models and techniques to detect real-timemalware injection on embedded processors using electromagnetic emission fingerprinting. Priorto the ML frenzy, Zach spent ten years as an embedded software engineer, developingoptimized algorithms for real-time computer vision, speech recognition, robotics, GPS, andinertial navigation systems.
Margaret Mayer is a Managing Vice President of Software Engineering for Messaging, ConversationalAI and Innovation Platforms at Capital One. She sets the technology strategy for Eno, CapitalOne’s proactive assistant, as well as enterprise consumer messaging capabilities for email,push and SMS. She is also accountable for emerging digital payments and Capital One Labs,which is the incubator for innovation. Margaret is also a passionate advocate for closing thegap in women in technology as a member of Capital One’s Women in Technology initiative.
Margaret has been with Capital One for 20 years, in roles of increasing responsibility withinTechnology and Operations. She holds a BS in Operations Research & Industrial Engineeringfrom Cornell University and her MS and PhD from Lehigh University in the same field. Prior tojoining Capital One, she spent three years as an Assistant Professor at the University ofVirginia, in Systems Engineering. Her research interests have been varied, but always alongthe common theme of using technology to solve complex business problems. Margaret is aboard member of CodeVA.org and the Computer Science Industrial Advisory Board at VirginiaCommonwealth University.