Abstract coming soon

We’ll examine the limitations of traditional rule-based systems in today’s complex fraud landscape and how AI is improving detection accuracy, scalability, and efficiency. The session will also provide a glimpse into AI models developed using Amazon Bedrock, highlighting real-world applications.

Key focus areas include the use of Large Language Models (LLMs) for suspicious activity detection, real-time monitoring to identify anomalies, and AI-driven decision-making to reduce manual review and intervention.

This session will offer practical insights into the evolving role of AI in fraud prevention, compliance automation, and risk management, while addressing both the opportunities and challenges of deploying intelligent systems in high-stakes environments.

This session explores how Generative AI can close that context gap. Drawing on experience from Nomain, we’ll show how AI-driven understanding of legacy systems enables faster analysis, safer transformation, and improved traceability — all while keeping developers in control of their tools and decisions.

By combining GenAI’s analytical power with human expertise, organisations can finally modernise mainframes without doing a “”big bang”” and still modernise at unforeseen speed. Because the future of mainframe modernisation isn’t about replacing the human — it’s about amplifying their insight.

Abstract coming soon

This presentation introduces a novel framework that combines statistical rigor, business logic, and Generative AI to deliver segments that are both analytically sound and operationally meaningful. The process begins by encoding institutional segment definitions—such as product mix, behavioral patterns, and risk appetite—into structured rules. Generative AI, powered by large language models and retrieval-augmented generation (RAG), then acts as a semantic evaluator, assessing cluster alignment with these definitions and key KPIs like profitability, churn, and credit risk.

By integrating semantic evaluation, KPI alignment, and traditional metrics into a single composite score, and feeding this back into the modeling process, the framework ensures explainable, business-aligned segmentation that meets regulatory expectations and drives strategic value.

Abstract coming soon

Abstract: This talk explores how S&P Global Market Intelligence applies machine learning and emerging AI techniques across the full fixed income pricing workflow—from ingesting raw market data to delivering high-quality prices for both liquid and illiquid securities. Our asset class coverage includes CDS, loans, and bonds. We will discuss how ML enhances upstream inputs such as document and quote parsing, enabling cleaner and more complete datasets for downstream pricing models. Within our evaluated pricing framework, ML supports curve construction, clustering of comparable securities, anomaly detection, and pattern recognition in sparse markets. We will also highlight how AI streamlines key operational processes, including validating price movements, generating explainability insights, and automating components of price-challenge responses. Together, these capabilities create a scalable, transparent, and adaptive pricing ecosystem that ensures high coverage and accuracy while meeting client expectations and supporting analyst oversight.

The evolution is especially critical in catastrophe modeling and pricing strategies, where the frequency and severity of natural disasters have rendered conventional actuarial methods less effective. Leading this innovation is me, leveraging cutting-edge AI to improve insurance pricing, streamline claims processes, and enhance risk assessment.

This session introduces and demystifies the LLM-as-a-Judge paradigm, a powerful and practical solution for automated LLM evaluation. I will discuss a production-ready path from rubrics to ROI by covering how to define criteria that matter for the product and choosing the right judging mode (single-output vs. pairwise). We will discuss how automated evaluation can accelerate iteration without losing fidelity to human expectations. Attendees will learn the foundational principles and practical steps for implementing this system, including constructing effective evaluation prompts, designing robust rubrics and scoring scales, and choosing the right “”judge”” model for the task.

The core focus of this presentation is to translate this technical capability into measurable business impact. The session will include demonstrating how implementing an LLM-as-a-Judge system drastically reduces evaluation time, enables faster product development cycles, and ensures the deployment of higher-quality, more reliable LLMs. The session will provide a clear framework for connecting evaluation scores to key business metrics like user satisfaction, customer support efficiency, or content generation quality, helping you to confidently demonstrate the return on investment of your LLM initiatives.
The talk will be grounded in recent community guidance on what works and what doesn’t, so teams can adopt judges confidently and avoid common traps. By the end of this talk, the attendees will have a clear understanding of the methodology and the confidence to implement it in their own organization to drive meaningful business results.