Generative AI systems are evolving beyond single-shot prompts toward closed-loop reasoning, a structured process of planning, acting, observing, reflecting, and refining until a verifiable goal is achieved. This session takes a practical, production-focused look at how to build such real-time reasoning agents from first principles.

Attendees will learn how to wire planning, memory, and tool use into adaptive reasoning loops that keep agents focused and reliable. The session compares frameworks such as BAML and PydanticAI, demonstrating tool calling (for APIs, retrieval, and structured outputs), short- and long-term memory management, and guardrails for control and observability. A live mini-demo will show an agent that plans research steps, calls tools, inspects evidence, revises its plan, and produces a structured, verifiable output, all in real time.

What You Will Learn

• How to design a THINK → ACT → THINK reasoning loop with clear state, termination, and fallback logic

• Strategies for managing short-term and long-term memory to balance efficiency and persistence

• How to evaluate and select orchestration frameworks for real-time reasoning and production readiness

Who Should Attend

AI engineers, software architects, and developers building reliable, adaptive agents that perform multi-step reasoning, data synthesis, and goal-directed problem solving.