DSS AUSTIN | Schedule

As AI adoption grows, relying on a single LLM for data extraction can lead to inconsistency and bias. This session explores how using multiple LLMs in parallel—combined with an automated LLM-as-a-Judge framework—can boost accuracy, reduce bias, and improve trust in extraction pipelines. Learn how to build scalable, resilient systems for unstructured data extraction without heavy human oversight.

Abstract coming soon

This session is intended for technology leaders, engineers, and architects involved in designing or delivering Agentic AI systems at enterprise scale. It examines why Agentic AI breaks in production, focusing on architectural boundaries, governance, controlled tool execution, failure isolation, and operational ownership.

Attendees will gain practical insight and proven design patterns to move Agentic AI systems from experimental demos to reliable, secure, and scalable enterprise platforms.

Drawing from real production rollouts, we’ll unpack key architecture patterns—from task decomposition and tool orchestration to memory and routing logic. You’ll learn how to move past brittle prompt chains, implement robust validation layers, and design agents that deliver measurable outcomes while aligning with enterprise governance and cost constraints.

Perfect for technical leaders and practitioners ready to operationalize AI agents at scale.

We’ll explore how to use Prompt Chaining, Semantic Routing, and Evaluator-Optimizer loops to design agents that are accurate, composable, and embedded into real systems. You’ll leave with a clear taxonomy of agent types and a practical roadmap for architecting AI systems that drive measurable ROI.

Drawing from experience across banks and high-growth startups, we’ll discuss how to operationalize global standards (EU AI Act, NIST AI RMF, OSFI E-23) to build trust with regulators, investors, and customers. The goal: make governance a catalyst, not a constraint.

This session dives into how a legacy Avaya IVR system was reimagined into an agentic, self-optimizing platform powered by Amazon Connect and RAG pipelines. Learn how we built SSML models that adjust tone and phrasing in real time, reduced AHT, and introduced an orchestration layer that updates strategies in minutes—not weeks.

If you’re scaling GenAI, modernizing CX, or looking to deploy agentic architectures in enterprise environments, this is your blueprint.

Learn how techniques like UEBA, unsupervised learning, and AI-driven threat detection are being used to detect insider threats, stop credential misuse, and build resilient, compliant frameworks for modern healthcare systems.

If you’re working at the intersection of healthcare, AI, and data protection—this session is for you.

Traditional rule-based notification systems fall short at scale. In this session, Preetham Reddy Kaukuntla shares how Glassdoor rebuilt its email and push infrastructure using ML models for both content and timing optimization. By integrating transformer-based subject line generation, LSTM and Prophet models for send-time prediction, and uplift modeling, Glassdoor now delivers 32M+ daily messages with measurable engagement gains. The session covers architecture, feature engineering, and lessons learned on model evaluation and long-term user impact.