Case Study — Week 3 — Mastering RAG · Advanced Agentic AI Tools

Advanced Agentic AI Tools

Overview

Week 01 Foundations of Agentic AI

Agentic AI vs Generative AI: Comparing Autonomy, Workflows, and Use Cases
The Agentic AI Operating Model
Persistent Instructions, Skills, Reverse Prompting, and Prompt Contracts
Multi-Agent Orchestration, Model Routing, Video-to-Action, and Parallel Browser Work
Multi-Agent Consensus, Debate, and Verification
Context Engineering, Model Routing, Cost Control, and Scaling

Case Study

Week 02 Mastering the Model Context Protocol (MCP)

Why MCP Exists
The Host-Client-Server Architecture
Tools, Resources, and Prompts
Communication Lifecycle, JSON-RPC, and Transports
Connecting to and Using an Existing MCP Server
Building a No-Code MCP Server
Building a Code-Based MCP Server
Development Environment, SDKs, and the Build Sequence
Testing and Debugging MCP Implementations
Authentication, Authorization, and Permissions
Prompt Injection, Tool Poisoning, and Dynamic Tool Changes
Supply-Chain Security and Secure Operations
Designing Reliable MCP Tools and Workflows
Deployment, Scalability, and Enterprise Infrastructure
Agentic, Multimodal, and Real-Time Applications
Applied MCP Case Studies
A Hands-on Learning Path
Community, Packaging, and Contribution
Ecosystem Directions and Advanced Projects
MCP Mastery Assessment

Case Study

Week 03 Mastering RAG

RAG Foundations and the Shift to Agentic Retrieval
End-to-End Architecture for a Multi-User RAG Application
Building with AI Agents - Planning, Context, Validation, and Version Control
From Application Shell to Managed RAG - Fast Start and Black-Box Limits
Owning the Retrieval Pipeline - Ingestion, Embeddings, Configuration, and Security
Maintaining Knowledge Quality - Deduplication, Metadata, Parsing, and Performance
Improving Retrieval Quality - Hybrid Search, Filtering, Reranking, and Tool Loops
Retrieving Beyond Documents - Text-to-SQL, Web Search, and Database-Level Control
Whole-Document Reasoning with Subagents and a Transparent Streaming Interface
Testing, Debugging, Local Operation, and Production Readiness

Case Study

Week 04 Memory Layer for AI Agents

Why Conventional RAG Is Not Enough for Stateful AI Agents
Represent Memory with Relational, Vector, and Graph Layers
Turn Raw Data Into Searchable, Structured Memory
Manage the Complete Memory Lifecycle
Partition Memory by Domain, Session, and Agent Role
Combine Vector Retrieval with Graph Traversal for Multi-Hop Answers
Use Feedback to Improve Future Retrieval and Decision Making
Orchestrate Multiple Agents to Resolve a Cross-System Incident
Design for Persistence, Observability, Latency, and Deployment
Understand the Trade-Offs and Choose the Right Memory Use Case

Case Study

Week 05 Mastering the Essentials of Kimi K3

Start with a Real Workflow Problem
Match the Model, Context, and Interface to the Workload
Design Goal Loops Instead of Chasing Perfect Prompts
Decompose Large Work Into Specialized Subagents
Turn Successful Procedures Into Reusable Skills
Engineer Context Continuity with Projects and Shared Memory
Extend the Agent with Tools, But Control the Trust Boundary
Deploy Recurring Work Where It Can Actually Run
Make Validation and Recovery Part of the Build Loop
Separate the Model from the Harness and Compose an Agent Operating System

Case Study

Week 06 Mastering the Essentials of Agentic Engineering

The Agent Harness Is the Operating Layer
Compress Context and Layer Instructions
Turn Every Run Into Persistent Learning
Parallelize Independent Work and Synthesize the Results
Use Stochastic Consensus and Debate to Explore the Solution Space
Build Lean Specialist Pipelines with Fresh Review Contexts
Design Autonomous Improvement as a Measured Experiment Loop
Choose the Least Complex Automation Layer That Works
Treat Workspace Organization as Part of Context Engineering
Contain Security Risk with Simple Controls and Independent Review

Case Study

Week 07 Mastering Git & GitHub

Version Control, Git, and GitHub
The Git Workflow and Repository Setup
Staging Changes with Control
Commits, Identity, History, and Comparison
Undoing, Restoring, Removing, and Reversing Changes
Branching for Isolated Development
Merging Branches and Resolving Conflicts
Remote Synchronization and Pull Requests
Preserving Unfinished Work with Git Stash
Rebasing for a Linear History

Case Study

Week 08 Mastering ComfyUI

Building a Local ComfyUI Environment and Model Foundation
Understanding the Node Graph and the Text-to-Image Pipeline
Controlling Generation with Resolution, Seeds, Steps, Cfg, Samplers, and Denoise
Organizing, Reusing, and Debugging Workflows
Extending ComfyUI with Manager, Models, and Custom Nodes
Upscaling Images and Controlling Which Outputs Continue
Using Image-to-Image to Balance Preservation and Change
Producing Detailed Enlargements with Tiled Image-to-Image Upscaling
Controlling Pose and Composition with ControlNet
Integrating External AI Tools and Consistent-Face Workflows with InstantID

Case Study

Week 09 Detailed Workflow for AI Coding

Manage the LLM Context as a Limited Engineering Resource
Reach a Shared Design Concept Before Requesting a Plan
Create a Destination Document Without Turning It Into a Specification-to-Code Contract
Turn the Destination Into Vertical-Slice Issues and a Dependency Graph
Separate Human Planning from AFK Implementation
Use Test-Driven Development and Feedback Loops as the Agent's Control System
Review in a Fresh Context and Use Manual QA to Restore Human Judgment
Shape the Codebase Around Deep Modules with Simple Interfaces
Control Instructions with Push and Pull, and Prevent Documentation Rot
Scale the Workflow with Planners, Isolated Implementers, Reviewers, and Mergers

Case Study

Week 10 Setting Up an LLM Council

Use an LLM Council for Consequential, Ambiguous Decisions
Build the Decision Process Around Stochastic Consensus
Create Perspective Diversity with Specialized Agent Prompts
Install and Register the LLM Council Skill
Operate the Default Single-Model Council
Add Model Diversity with a Cross-Vendor Council
Configure Cross-Vendor Access and Protect the API Key
Run and Inspect the Cross-Vendor Workflow
Choose Between Default and Cross-Vendor Mode

Case Study

Week 11 Advanced AI Engineering Workflows

Move from Tactical Programming to Strategic Programming
Optimize the Harness, Not Only the Model
Design Skills as Reusable Procedures Without Bloated Context
Use Adversarial Interviewing Before Implementation
Build Stateful, Mission-Centered AI Learning Workflows
Treat Human Knowledge, Skill, and Wisdom as the Performance Ceiling
Run Scoped Agents Away from the Keyboard Inside Sandboxes
Orchestrate Agent Work with Queues, Not Endless Loops
Move Human Checkpoints Rightward While Preserving Review and Observability
Keep Product Vision and Customer Discovery Human-Owned

Case Study

Week 12 How-To Guide: Hermes Agent

Distinguish the Model from the Hermes Agentic Harness
Design Hermes as a Self-Improving AI Employee
Choose a Runtime, Cost Model, and Recovery Path
Grant Computer Access Without Surrendering Control
Build a Multi-Channel Operating Layer
Shape Behavior with Identity Files and Profiles
Engineer Durable, Time-Aware Memory with Honcho
Add Observability and Deterministic Project Tracking
Scale from One Agent to Profiles, Squads, and Model Routing
Connect External Applications Through MCPs, CLIs, and APIs
Turn Repeated Work Into Cron Jobs, Skills, and Agentic Workflows
Use /Goal for Long-Horizon, Self-Checking Execution
Assemble a Visual Creation and Deployment Stack
Build a Content Production and Distribution Engine
Create Personalized Lead-Generation and Property Marketing Workflows
Operate Community Management and Customer Support Agents
Build a Paper-Trading Agent with Separate Execution and Strategy Layers
Build a Private Health Accountability System
Create a Jarvis-Style Personal Command Center
Productize Hermes Systems as Outcome-Based Services

Case Study

Course › Week 3 › Case Study

Case Study

Week 3 — Mastering RAG

This hands-on case study is being prepared and will be added soon.

Scenario brief
Your tasks
Submission & review

Coming soon

Advanced Agentic AI Tools — 12-week learning module