Embedded Files
Head of Data, Grammarly
Eric is currently the Head of Data at Grammarly, leading teams of data scientists and software engineers who are responsible for improving how teams measure and iterate on the product to create the best communication assistant at Grammarly. He has over 15 years of experience in data leadership roles at Grammarly, Stitch Fix, Yelp, LinkedIn, and CoreLogic.
He is also passionate about sharing knowledge and insights on data product management and data science as a product. He writes a newsletter called "From Data to Product". His mission is to help data professionals and product managers build better data products and use data to help companies assess and navigate risk and opportunity. Eric has a PhD and Masters in Mathematics from Arizona State University, an MS in Business Analytics from University of Minnesota, and an MBA from University of Chicago-Booth.
Title: Understanding What's Not Measured: Closing Gaps in the End-to-End Customer Journey
Abstract: A common pitfall in defining the end-to-end customer journey is using a bottom-up approach of existing metrics and instrumentation to piece together a view of how customers find, engage with, and retain within a product. This approach can expose significant gaps in understanding the customer that may not be obvious unless there is prior art about what "good" looks like. In this talk, I'll explore some common "misses" in measuring the end-to-end journey and how to close them.
Associate Professor, Department of Computer science and Engineering, University at Buffalo
Changyou Chen is an Associate Professor in the Department of Computer Science and Engineering at the University at Buffalo, State University of New York. He is interested in general AI techniques, with current research focusing on the topics of multi-modal modeling and learning, self-supervised learning, large language models, generative models, large-scale Bayesian sampling and inference. Previously, he was an Assistant Professor at the University at Buffalo, a Research Assistant Professor and a Postdoctoral Associate in the Department of Electrical and Computer Engineering at Duke University. He earned his PhD from College of Engineering and Computer Science, the Australian National University. He has co-authored 200+ papers on machine learning and AI. He serves as an Area Chair for multiple AI conferences including ICML, NeurIPS, AAAI, IJCAI, and an Action Editor for Transactions on Machine Learning Research.
Title: Behavior-Driven Content Intelligence: A Paradigm for Customer Journey Optimization
Abstract: This talk presents our work on integrating behavior data into modern AI models. Modern AI models such as generative models excel at creating visually impressive content, yet this often fails to translate into effective business outcomes like user engagement and conversion. This talk introduces Behavior-Driven Content Intelligence, a framework that aligns generative AI with tangible business goals by using real-world user behavior as the primary training signal. We demonstrate how a suite of models can learn from behavioral data (e.g., likes, clicks, shares) to predict content performance, generate highly engaging visuals, and simulate user reactions through modern LLMs, which creates a continuous learning loop that makes content provably more effective. Our research offers a new paradigm for leveraging AI to optimize the entire customer journey.
Staff Data Engineer at Spotify
Will gets to explore data and build new and novel datasets to help understand user behavior and app features at Spotify. Will has over 35 years in the industry. The early years were spent building the yearly smartphones and the later years doing things like real-time telecom billing. These days he dabbles in data science.
Title: Using small language models to reveal user patterns and preferences from behavioral timelines
Abstract: A walk-through of training models on user journeys to predict interesting events. The example investigation will use an LLM to isolate a high-friction journey that has high likelihood of user churn.
Staff Research Scientist, Netflix
Yesu Feng is a Staff Research Scientist and tech lead on the Foundation Models for Personalization team at Netflix. He leads the development of large-scale foundation models—including both behavior-driven representation models and language models—to power personalized search and recommendation across the Netflix product. His work focuses on integrating self-supervised learning and generative modeling into real-world personalization systems, helping connect millions of members with content tailored to their tastes and interests. Prior to Netflix, he worked on feed recommendation at LinkedIn and supply-side modeling at Uber.
Title: From Rows to Dialogues: GenAI for Homepage and Search Personalization
Abstract: Netflix personalizes the user journey through two key canvases: the homepage and search. This talk presents how generative AI transforms both. On the homepage, a foundation model unifies user representation and powers personalized ranking across tasks. In search, GenSearch introduces a conversational interface driven by fine-tuned LLMs, enabling natural, multi-turn interactions grounded in user context. Together, these systems illustrate a cohesive GenAI strategy to guide members from inspiration to discovery.
Page updated
Google Sites
Report abuse