Invited Talk

Vanja Josifovski, NVIDIA

Bio: Vanja Josifovski is a computer scientist and technology executive renowned for his contributions to artificial intelligence, machine learning, and search technologies. He is best known as the co-founder and CEO of Kumo.AI, a predictive AI platform for enterprises founded in 2021. Previously, Josifovski served as Chief Technology Officer of Homes at Airbnb starting in 2019, where he led engineering, data science, marketplace, personalization, and regulatory teams, and before that as CTO and Vice President of Engineering at Pinterest, overseeing the company's technical vision, strategy, and talent development

Title: Predictions on relational data without predictive modeling

Abstract: In this talk, Vanja will present KumoRFM, the first Foundation Model for relational enterprise data, designed to perform zero-shot predictions directly on structured, multi-table data. Built on a scalable Graph Transformer architecture, KumoRFM generalizes across schemas and tasks, eliminating the need for hand-crafted features, data pipelines, or model training. Developers can try KumoRFM with their own data—no feature engineering required. By transforming relational data into heterogeneous graphs and applying in-context learning, the model delivers high-accuracy predictions across a wide range of applications—from churn and fraud detection to personalized recommendations and forecasting. We'll share key technical design choices, pretraining methodology, and lessons learned from deploying KumoRFM in real-world enterprise environments. This session highlights how the architecture enables fast time-to-value for AI initiatives while maintaining model transparency and performance. Attendees will gain insights into the future of predictive modeling, where structured data can be leveraged with foundation models as directly and flexibly as text or images.

Invited Talk

Paul Bennet, SPOTIFY

Bio: Bio: Paul Bennett is Director of Research for Generative AI at Spotify, where his research focuses on adapting and aligning large language models to support creators and audiences in responsible, human-centered ways. His work spans domain adaptation of generative models, preference alignment, and scalable evaluation frameworks that enable models to operate reliably under real-world ambiguity. Paul’s recent research explores how to teach generative systems to reason in structured domains through representations like semantic identifiers, optimize continuously to reflect nuanced human preferences, and generate contextualized narratives that enhance user experience. Across this work he studies how model design and evaluation choices influence alignment with user and creator intent, safety, and controllability. Prior to joining Spotify, Paul was a Partner Research Manager in the Augmented Learning and Reasoning group at Microsoft Research. His research has been published broadly in machine learning and human-centered computing venues and recognized with multiple awards, including a SIGIR Test of Time Award. His work reflects a commitment to advancing generative AI research that bridges technical depth with human-centered design and safety.

Title: TBD

Abstract: TBD

Invited Talk

Baoyuan Wang, ZILLOW

Bio: Baoyuan Wang is a Distinguished Scientist at Zillow, where he leads AI CoPilot development to enhance the home-shopping experience. His work spans multimodal representation learning, vision and NLP, generative AI, large language models, and agentic systems that integrate perception, reasoning, and action. Before joining Zillow, he led AI Companion Core at Zoom, co-founded Xiaobing.ai and oversaw multimodal R&D, and spent more than a decade at Microsoft across Microsoft Research, HoloLens, Azure AI, and Xbox Kinect. He received his BS and PhD from Zhejiang University and has published widely in AI.

Title: From Search to Guidance: Reimagining Real Estate Marketplace with Agentic AI

Abstract: Real estate is among the most financially significant and emotionally complex decisions people make, yet most digital experiences remain centered on information retrieval rather than intelligent guidance. In this talk, I'll share Zillow's journey from building high-IQ systems that answer questions to developing agentic AI with both AQ (Agentic Quotient) - the ability to reason, plan, and act - and EQ (Emotional Intelligence) to support users throughout their home-buying and selling journeys in a real estate marketplace.

A central argument of this talk is that strong AQ requires more than a capable foundation model. It depends on a domain-specific world model that captures the rich relationships among homes, neighborhoods, markets, users, and transaction workflows. I'll discuss how we combine this panoptic understanding with reasoning, planning, memory, and tool-use to build AI copilots that proactively guide users instead of simply responding to requests. Rather than presenting a single system, this talk focuses on the emerging problem framing of agentic AI for high-stakes domains. I'll discuss the architectural principles behind deploying these systems in production, the challenges of grounding, orchestration, trust, personalization, and evaluation, and a set of open research questions around building trustworthy agents that understand, reason about, and act within complex real-world environments