Nội dung:

• LLM Fundamentals

• Transformer & Attention

• Prompt Engineering

• Tool Calling

• MCP (Model Context Protocol)

• AI Agent Landscape

• Lab: Ví dụ điển hình về Agentic: Claude Code

Context Harness

• RAG Architecture

• Chunking Strategies

• Embedding Models

• Reranking

• Hybrid Search

• Agent Memory

• Context Compression

• Long Context Optimization

Lab

• Xây Enterprise RAG

• Multi-source Knowledge Base

• Docker là gì?

• Phân biệt image và container

• Cài đặt Docker

• Xây dựng Docker Image và chạy docker Container

• Docker container

Tool Harness

• Function Calling

• MCP Servers

• API Integration

• Tool Registry

• Least Privilege Design

• Retry Mechanisms

• Error Recovery

• Audit Logging

Lab

• CRM Agent

• ERP Agent

• Database Agent

• Lập trình reasoning

• Lập trình gọi tools

• Lập trình truy xuất memory

Hướng dẫn làm bài tập

Orchestration Harness

• Agent Planning

• DAG Workflows

• State Machines

• LangGraph

• Multi-Agent Systems

• Routing Strategies

• Fallback Models

• Parallel Execution

Lab

• Research Agent

• Customer Service Agent

• Why Multi-Agent?

• Single Agent Limitations

• What is Multi-Agent?

• Architecture

• Agent Roles

• Collaboration Patterns

• Planning & Orchestration

• Communication

• Memory

• Parallel Execution

• Human-in-the-loop

• Frameworks

• Demo

• Best Practices

Evaluation Harness

• Golden Datasets

• LLM-as-a-Judge

• Human Feedback

• A/B Testing

• Regression Testing

• RAG Evaluation

• Hallucination Detection

Lab

• Build Evaluation Pipeline

AgentOps Harness

• Monitoring

• Tracing

• Cost Tracking

• Latency Optimization

• Accuracy Metrics

• Drift Detection

• Observability

Tools:

• LangSmith

• OpenTelemetry

• Phoenix

• Datadog

Security Harness

• Prompt Injection

• Jailbreak Defense

• Data Leakage Prevention

• PII Protection

• Access Control

• Credential Management

• Secure Tool Access

Lab

• Attack & Defend AI Agents

Governance Harness

• Audit Trail

• Explainability

• Approval Workflows

• Human-in-the-loop

• AI Policies

• Risk Management

• Compliance Frameworks

Lab

• AI Governance Dashboard

Xây dựng một hệ thống hoàn chỉnh gồm:

• Knowledge Base

• RAG Pipeline

• AI Agents

• Tool Calling

• Workflow Engine

• Evaluation System

• Security Layer

• Governance Layer

• Monitoring Layer