DSS SF | Schedule

The integration of Generative AI (GenAI) and MLOps is transforming the way organizations measure and optimize marketing performance. This session will explore how GenAI can be leveraged to extract actionable insights from complex Bayesian hierarchical Marketing Mix Models (MMM) and Incrementality Testing Systems, enhancing the interpretation of model outputs and driving business decisions. By automating the generation of narratives from sophisticated data models, GenAI ensures that technical insights are translated into clear, business-relevant summaries that can be easily understood by executives.
At the same time, MLOps plays a crucial role in the scalability and operationalization of these advanced measurement systems. It supports continuous learning, automated model retraining, and the deployment of robust, enterprise-grade marketing measurement frameworks on Databricks that ensure transparency, reproducibility, and governance at scale. Together, GenAI and MLOps streamline the process of model development and execution, providing organizations with a scalable and agile system for marketing data science.
This session will discuss real-world examples and the challenges of implementing MMM and Incrementality Testing at scale. Attendees will gain insight into the best practices and design patterns that bridge the gap between marketing data science and business strategy, empowering marketers to make faster, data-driven decisions.

As agentic AI systems transition from prototypes to real-world clinical use, data scientists must balance autonomy with trust, safety, and compliance. This talk highlights the unique evaluation and monitoring challenges of agentic AI in healthcare:

• Evaluation gaps: why conventional benchmarks miss multi-step reasoning, tool use, and localized workflows—and how contamination and fragile metrics distort results.
• Bias and safety: demographic bias and other risks that trigger regulatory, legal, and contractual obligations for safety and fairness assessments.
• Continuous monitoring: practical MLOps strategies for detecting drift, unsafe autonomy, and compliance issues in deployed systems.
• Tools and standards: open-source libraries like LangTest and MedHELM, new stress-test datasets such as MedSafetyBench and MedAgentBench, and emerging guidance from NIST and CHAI.

The lessons shared here will not only help teams working on clinical AI, but also anyone building agentic systems in highly regulated industries where safety, fairness, and compliance are paramount.

Abstract comin soon

Using a real-world GSX case study, Matt demonstrates how Genesis agents on Snowflake automate every stage of the data lifecycle — from mapping to monitoring — while maintaining enterprise-grade guardrails for governance and compliance. The result: delivery cycles cut from 3 months to 5 hours and 400+ engineering hours recovered each month. The session concludes with a live demo and a new north star for data leaders — Acceleration as a KPI.

Abstract cooming soon

In an era of information overload, helping users find exactly what they need—quickly and effortlessly—has become a cornerstone of great product design. At the heart of this lies the art and science of personalized recommendations, now accelerated by the power of Generative AI (GenAI) and Machine Learning. This session explores how AI transforms broad, generic search queries into personalized, visually intuitive experiences using token-based refinement systems. For example, a user searching for “laptop” is guided through a GenAI-powered experience that refines their query into tailored, meaningful options such as gaming, lightweight, touchscreen, or under $1,000. We’ll explore how AI is applied at every stage of this journey, from token generation and normalization to data merging, diversity enforcement, and dynamic image generation. The session will also highlight how these innovations positively impact downstream metrics like CTR, conversion, and engagement. Join us to see how LLM-driven logic, ranking models, and NLP are redefining the way users search, discover, and shop.

Large language models (LLMs) like GPT-4 are revolutionizing the field of Information Retrieval (IR) by enhancing traditional search methods with generative AI capabilities. This session explores how LLMs are reshaping search systems, moving beyond keyword-based retrieval to contextually aware, natural language-driven search. By leveraging LLMs, search engines can generate more accurate, relevant, and coherent results tailored to user intent, transforming the search experience. The integration of generative AI allows for nuanced understanding of complex queries and provides users with comprehensive, human-like answers. This session will delve into the architecture of LLMs, their application in modern search systems, and how they enhance retrieval with techniques like retrieval-augmented generation (RAG). We will also address the challenges and opportunities of incorporating LLMs in large-scale search systems, including issues of scalability, efficiency, and response accuracy.

Abstract coming soon

This talk explores the rapidly evolving role of AI agents in the forecasting world and whether these new capabilities are poised to replace traditional ML models. This talk will also cover how they can work in conjunction with existing systems to unlock new efficiencies, and the revolutionary implications for data scientists and business operations.

Most personalization systems rely on surface-level correlations, but correlation doesn’t mean causation. To truly understand what drives user engagement, conversions, or retention, we need a shift toward causal modeling.
In this session, I’ll share how we applied causal inference techniques at Sling (EchoStar) to elevate our personalization strategy. We’ll cover:
A lightweight framework using uplift modeling and causal graphs
Practical lessons from moving beyond A/B testing into causal effect measurement
Real-world results using counterfactual analysis to improve recommendations
Attendees will leave with actionable methods for identifying causal drivers in user behavior and reducing noise in personalization metrics. This talk is ideal for data scientists, ML engineers, and product leaders working in consumer tech, streaming, or any data-rich environment.

When AI is deployed in healthcare, the stakes aren’t just high—they’re human. In this talk, I’ll share how we built and deployed trustworthy, scalable AI systems to automate critical decision-making processes like prior authorization, fraud detection, and risk prediction at Elevance Health, one of the nation’s largest healthcare providers. The work blends generative AI, causal inference, neurosymbolic reasoning, and adaptive dashboards to support medical and policy decisions where transparency, accuracy, and compliance are non-negotiable. Rather than chasing model performance alone, we designed AI that clinicians and policy teams could trust—and actually use. I’ll walk through how we identified real-world bottlenecks, embedded explainability into our systems, ran controlled A/B experiments to measure adoption, and reimagined workflows with self-learning components. This talk is for data scientists and leaders building AI in regulated or high-impact environments, where success is not just about prediction but about credibility, integration, and long-term value.