AgentStack (MeetUp)

× ×

This talk will present the fundamental concepts of AI agents and highlight key challenges in their development and deployment. We will also explore strategies to enhance agentic workflows, such as self-refinement and debating techniques, to improve their performance

× ×

Prototyping AI agents is one thing—ensuring they are robust, reliable, and production-ready is another. As AI agents take on increasingly complex tasks, from autonomous decision-making to tool execution, traditional evaluation methods fall short, lacking scalability and the ability to capture real-world multi-step reasoning and dynamic interactions.

In this talk, I will introduce IntellAgent, an open-source multi-agent framework designed to rigorously test AI agents in diverse, high-stakes environments. IntellAgent generates real-world scenarios, simulates complex user interactions, and systematically evaluates policy adherence, reasoning accuracy, and failure patterns. By leveraging policy graphs, complexity-driven test case generation, and automated critique mechanisms, it helps developers identify weaknesses, optimize agent performance, and prevent failure modes before deployment.

As AI systems become more autonomous and deeply integrated into critical workflows, reliable evaluation is no longer optional—it’s essential. IntellAgent provides the missing evaluation layer, offering a structured, scalable testing framework that ensures AI agents perform as expected before real-world deployment, setting new reliability standards for AI-driven systems.

× ×