AI Con USA 2026 - AI Engineer

Enterprise containerized applications face a critical reliability crisis with complex failure modes including memory leaks, cascading failures, network partitions, and resource contention that traditional monitoring tools cannot predict or resolve fast enough. Organizations typically experience multiple production incidents monthly with multi-hour resolution times that consume significant engineering resources while causing customer-facing outages and revenue loss. Traditional approaches rely on reactive monitoring, manual troubleshooting across distributed container environments, and time...

AI agents have evolved beyond LLM chatbots; they possess the ability to plan, reason, and act autonomously. However, as their autonomy increases, understanding how they make decisions becomes more challenging. Traditional methods of observability—such as metrics, logs, and traces—capture outcomes but do not reveal the underlying reasoning. This session will explore how AI Agent Observability can shed light on the decision-making process by collecting and analyzing agent traces. We will discuss emerging standards like the Model Context Protocol (MCP), which provides structured and shareable...

Imagine starting your workday to find all your company’s sensitive data has been leaked due to a nearly imperceptible hack embedded in an AI prompt. The culprit isn’t the AI model but rather a lack of protections surrounding it. When rapidly adopting AI at scale, many organizations unintentionally ignore a crucial element: security and governance. Often, companies will only add operational AI rules post-deployment. However, this approach creates hidden blind spots and slows security teams’ responses when threats inevitably appear. In this presentation, Mark Toler will reveal why AI...

Are you ready to transform your productivity from incremental gains to exponential growth? Whether you're an individual contributor or engineering leader, this session introduces the AI Maturity Model, a proven framework that guides technical teams through five distinct levels of AI adoption—from 33% productivity boosts to an extraordinary 1000% increase. Discover how to navigate each stage: Level 1 (Foundation, 33%) builds essential AI awareness; Level 2 (Literacy, 75%) develops practical AI skills; Level 3 (Fluency, 300%) masters AI-assisted workflows; Level 4 (Agents, 500%) implements...

As Large Language Models (LLMs) transition from conversational tools to autonomous AI agents capable of high-stakes, multi-step actions, the reliance on generalized, default safety alignment from model providers has become insufficient and often hazardous. This session addresses the critical need for application-specific, custom guardrails necessitated by persistent jailbreak vulnerabilities, the high cost of achieving effective fine-tuning alignment, and the challenge of enforcing proprietary organizational compliance (e.g., contextual PII protection or specific regulatory mandates unique...

As AI models grow in complexity and scale, inference efficiency has emerged as a critical engineering challenge for enterprise deployment. Traditional infrastructure built for training workloads often fails to meet the latency, throughput, and cost demands of large-scale inference operations. In this session, Sandeep will be sharing practical insights from engineering AI infrastructure at Broadcom, focusing on the end-to-end optimization of compute, networking, and storage subsystems. The talk explores techniques such as dynamic workload placement, adaptive batching, model quantization,...

Traditional software testing is fundamentally deterministic: the same inputs must always produce the same outputs. Yet many teams introduce AI into their testing without first defining the problem the AI is meant to solve, leading to brute-force experimentation and unreliable results. Google’s 2025 DORA report highlights that user-centricity is a prerequisite for AI success and that AI is most effective when it is pointed at a clear problem. AP shows you how that insight applies technically to testing. Before AI can be used as a testing tool, it must first be tested and understood in the...