Agentic System Design - Agentic System Design

Search⌘ K

AI Features

Agentic System Design

1.Agent Design Fundamentals

Introduction to AI Agents Agent Architecture: Core Agent Components Agent Architecture: Components Interaction and Agent Memory Structuring Agent Behavior: Agent Orchestration Patterns Building Trustworthy Agents: Guardrails and Human Oversight Key Challenges and Design Strategies in Agentic AI Systems

2.Multi-Agent Conversational Recommender System (MACRS)

Introduction to MACRS and the Design Challenge MACRS Multi-Agent Act Planning Framework MACRS User Feedback-Aware Reflection Mechanism Evaluating MACRS: Performance and Insights

Breakout Session

Design Your First Agent

3.Nvidia Eureka Learning Agent

Eureka: Automating Reward Design with Coding LLMs Eureka’s Zero-Shot Reward Generation Eureka’s Evolutionary Search for Iterative Improvement Eureka's Reward Reflection Mechanism Eureka from Human Feedback and Novel Reward Discovery Evaluating Eureka: Performance and Insights

4.Implementing a Eureka-Like Reward Learning Agent with Google ADK

Agent Design, Code Structure, and Output Demonstration SetupAgent and Environment Grounding Reward Generation and Evaluation Loop Selection, Reflection, and Human Feedback Loop Control, Exit Conditions, and System Behavior

Breakout Session

Thinking Through Eureka

5.Applying Agentic Design Principles

Thought Exercise: Design a Smart Parking Agent System

6.Designing an AI Agent for Generating LLM Pipelines

Introduction to ChainBuddy and the “Blank Page Problem”The Requirement-Gathering Chat Assistant The Multi-Agent Pipeline Generation Framework Evaluating ChainBuddy: Performance, Usability, and Design Insight

7. Designing a Web Agent

The Multimodal Web Agent Challenge WebVoyager’s Architecture: A Multimodal ReAct Loop Text vs. Multimodal Web Agents in a Real-World Task Evaluating WebVoyager and Design Insights Thought Exercise: Designing a Self-Improving Web Agent

8.Designing a Multimodal-LLM Agent for Multi-Object Diffusion

Introduction to MuLan and the Multi-Object Generation Challenge MuLan’s Planning and Progressive Generation VLM-Feedback Control and Human-in-the-Loop Interaction Evaluating MuLan: Performance and Design Insights

9.Thought Exercise: AI Hospital

Design a Multi-Agent Medical Diagnosis System

10.OpenClaw Design

Designing a Personal AI Assistant like OpenClaw

11.Wrapping up

Mock Interview

Agentic System Design