Antonio Gulli
Google CTO 办公室工程总监
.....
待定
待定
Agentic Design Patterns：A hands-on Guide to Builing
Intelligent System （拟定）
Abstract: The leap from raw reasoning engines to production-ready autonomous systems requires a fundamental shift in software engineering. While LLMs provide the "brain," Agentic Design Patterns provide the skeletal structure necessary for reliability, scalability, and safety. Based on a body of work that has garnered over 1.4 million views on LinkedIn—with all proceeds from the resulting book dedicated to charity—this session codifies the twenty essential patterns required to move beyond simple chat interfaces. We will explore the mechanics of:
Reliable Workflows: Prompt Chaining and Iterative Reflection.
Dynamic Execution: Tool Use, Planning, and Multi-Agent Orchestration.
Framework Integration: Practical implementation using LangChain, CrewAI, and beyond.
The presentation moves past the technical toolkit to advocate for "Compassionate Autonomy," a design philosophy centered on building AI with purpose and human-centric responsibility. We conclude by tackling the "frontier friction"—the critical research gaps that currently hinder mass deployment. We will analyze the mechanics of Cascading Reliability Failure, the necessity of Dynamic Guardrails, and the challenge of Autonomy Calibration. Attendees will walk away with a rigorous, pattern-driven roadmap for architecting the next generation of dependable AI agents.
Reliable Workflows: Prompt Chaining and Iterative Reflection.
Dynamic Execution: Tool Use, Planning, and Multi-Agent Orchestration.
Framework Integration: Practical implementation using LangChain, CrewAI, and beyond.
The presentation moves past the technical toolkit to advocate for "Compassionate Autonomy," a design philosophy centered on building AI with purpose and human-centric responsibility. We conclude by tackling the "frontier friction"—the critical research gaps that currently hinder mass deployment. We will analyze the mechanics of Cascading Reliability Failure, the necessity of Dynamic Guardrails, and the challenge of Autonomy Calibration. Attendees will walk away with a rigorous, pattern-driven roadmap for architecting the next generation of dependable AI agents.
从原始推理引擎迈向可落地投产的自主智能系统，软件工程层面需要一次根本性变革。大语言模型（LLM）为智能体赋予 “大脑”，而智能体设计模式则为系统搭建起保障可靠性、可扩展性与安全性的核心骨架。课程系统梳理了突破简易对话交互、构建高阶智能体必备的 20 种核心设计模式，将深入拆解三大核心技术体系的底层原理：
可靠工作流：提示词链式调用、迭代自省机制
动态执行能力：工具调用、任务规划、多智能体协同编排
框架工程落地：基于 LangChain、CrewAI 等主流框架的实战落地方法
本次分享不止局限于技术工具讲解，更倡导人文向善自主智能设计理念 —— 以使命价值为内核、以人本责任为准则构建人工智能。
最后，课程将直面阻碍 AI 智能体大规模落地的前沿瓶颈难题：深度剖析级联可靠性失效的内在机理、动态安全防护围栏的刚需价值，以及自主权限校准的行业挑战。参会者将收获一套严谨完备、以设计模式为驱动的路线图，用于架构打造下一代高可靠 AI 智能体。
可靠工作流：提示词链式调用、迭代自省机制
动态执行能力：工具调用、任务规划、多智能体协同编排
框架工程落地：基于 LangChain、CrewAI 等主流框架的实战落地方法
本次分享不止局限于技术工具讲解，更倡导人文向善自主智能设计理念 —— 以使命价值为内核、以人本责任为准则构建人工智能。
最后，课程将直面阻碍 AI 智能体大规模落地的前沿瓶颈难题：深度剖析级联可靠性失效的内在机理、动态安全防护围栏的刚需价值，以及自主权限校准的行业挑战。参会者将收获一套严谨完备、以设计模式为驱动的路线图，用于架构打造下一代高可靠 AI 智能体。
敬请期待
......
.....
待定
待定
敬请期待
....
关注QECon公众号
关注QECon视频号
议题投稿
lijie@qecon.net
票务联系
18649077637 Lily
媒体合作
135-1619-6409 皮皮
商务合作
151-2264-3988 木子
购票咨询
18649077637 Lily
