Mini Map
Search
⌘ K
AI Features
Log In
Agentic System Design*
1.
Agent Design Fundamentals
Understanding AI Agents
Agents: Decision-Making Systems with Purpose
Model vs. Agent: Reactive text generator vs. Autonomous decision
Chatbot vs. AI Agent: A Comparison
Understanding the Agent Loop
Improve Decision Making for Flight Booking
The Core Components of an AI Agent
Understanding the Model
Understanding the Power of Tools
Instructions for Agent's Behavior
Agent's Reasoning and Constraints
Why Memory Matters
Short-Term vs Long-Term Memory
Accessing External Knowledge
The Importance of Memory in Multi-Step Tasks
Understanding Orchestration
Highlighting the Power of Orchestration
Types of Orchestration
Tool-Calling Loop
ReAct Orchestration
Plan-and-Execute
Fixing Broken Weather API Agent
Troubleshooting Email Notification Issue
Create a Weekly Blog Schedule for SaaS Product Launch
Understanding Multi-Agent Systems
Highlight: Multiple Agents for Specialized and Reliable Decision-
Comparing Multi-Agent Coordination Patterns
Generate a Competitive Analysis for Product Launch
Multi-Agent Systems vs Single-Agent Designs
Common Reasons for Agent Failures
Recognizing Common Failure Patterns
Weak Reasoning: The Root of Most Failures
Failure Modes: When an Agent Relies on Tool Use
Safety Guardrails
Mitigating Risky Requests: Implementing Safety Guardrails
Guardrails and Human Oversight
Human Oversight Levels
Quiz: Oversight Check - Which oversight mode requires humans to a
Balancing Autonomy and Accountability in Agent Design
Agent Foundations: Chapter Summary
2.
Multi-Agent Conversational Recommender System (MACRS)
Understanding User Intent Evolution in Recommendation Conversatio
Challenges Faced by Single Agents
The Multi-Agent Approach
Comparison of Single vs Multi-Agent Recommenders
Design Goals of a Multi-Agent Recommender
Understanding the System Agents
Understanding the Planner Agent
The Explorer Agent
The Evaluator Agent
How Agents Improve Through Reflection
Improving User Recommendations for Entertainment Preferences
Specialized Agents: Deeper Preferences and Richer Options
3.
Nvidia Eureka Learning Agent
The Challenge of Reward Design
Challenges with Manual Rewards
Meet Eureka
Human vs AI-Generated Rewards
Understanding the System Workflow
Understanding Robot's Environment Context
Exploring Reward Variations
Reflective Refinement for Reward Alignment
Unintended Consequences of Reward Design
EUREKA-style Reward Optimization
Understanding System Evaluation
Quiz: Understanding the System
4.
Designing an AI Agent for Generating LLM Pipelines
Understanding LLM Pipelines
Understanding Four Pipeline Archetypes
Why Manual Pipelines Break Down
Comparison: Manual vs Agent-Designed Pipelines
Benefits of Using an Agent
How Agents Plan: Three Strategies
Understanding Pipeline Representation
Building Effective Pipelines with a Component Library
Ensuring Pipeline Data Integrity
Improving Conflict Handling in Multi-hop Reasoning Pipelines
Detecting Complex Pipeline Failures
Agents Refining Pipelines
Agent Task Ordering Problem: Clearing the Sequence
Pipeline Reasoning: Which approach helps an agent design a cohere
Effective Pipelines: Structured Reasoning Flows
Designing LLM Pipelines Summary
5.
Designing a Web Agent
Understanding Web Agents
Understanding Web Pages: DOM Analysis vs. Visual Interpretation
Understanding the Challenges of Web Agents
Understanding DOM-Based Agents
Understanding DOM Signals for Interactivity
Understanding DOM-Based Navigation
Understanding DOM Interaction Primitives
Understanding DOM-Based Action Validation
Understanding Common DOM Failure Modes
The Need for Multimodal Agents
Understanding Pages Through Screenshots
Visual Element Identification
Visual Grounding for Accuracy
Visual Action Decoding
Screenshot-Based Validation
Understanding Common Multimodal Failure Modes
When to Use DOM vs Visual Signals
Hybrid Navigation Planning
Hybrid Recovery Strategies
Reflection Across Modalities
Improving Web Scraping Accuracy
Improving Signup Form Completion
Agent Behavior: Misinterpreting Visual Cues
Highlight: Modern Web Agents Mimicking Human Web Navigation
6.
Designing a Multimodal-LLM Agent for Multi-Object Diffusion
Understanding Multi-Object Diffusion
Challenges in Multi-Object Scenes Generation
Role of a Multimodal LLM
Benefits of Using an Agent
Core Abilities of an Agent
The Multi-Object Generation Loop
Planning the Scene Layout
Guiding the Diffusion Model
Image Model Inconsistencies: Addressing Object Merging and Placem
Improving Object Generation Process
Improving Object Correction in Generated Scenes
Identifying Common Failure Modes
A Strong Multimodal Agent: Acting Like a Director
7.
Conclusion
Thinking Like an AI Architect: Designing Agents and Building Reli
Shaping LLMs into Autonomous Agents
From LLM User to Agent Designer
AI Automation: Bridging Human Goals with Machine Capabilities
Key Takeaways
Final Reflection: What Makes an Agent Trustworthy?
The Future Belongs to Designers
Course Recap: You Built the Foundations of Agentic System Design
Claim your Certificate
Home
Courses
Agentic System Design*
Reflective Refinement for Reward Alignment
We'll cover the following...
This component is not supported in the current context
...