Agentic AI Engineering with Python — Build Autonomous Agents Training Course

Core Concepts of Agentic AI

• Defining autonomous agents: key definitions and taxonomy
• The agent loop: the perceive, decide, act, and observe cycle
• Establishing design patterns for agent responsibilities and scope

Python Tooling and Agent SDKs

• Leveraging LangChain and comparable SDKs to initialize agents
• Async programming, task queues, and subprocess management
• Packaging, virtual environments, and ensuring reproducible development workflows

Integration of External Tools and APIs

• Crafting tool interfaces and secure invocation patterns
• Establishing connections to web APIs, databases, and internal services
• Managing credentials, secrets, and enforcing least-privilege access

Memory, State, and Context Management

• Short-term context windows and advanced prompt engineering techniques
• Long-term memory architectures: Redis, vector stores, and retrieval augmentation
• Ensuring consistency, implementing caching strategies, and maintaining memory hygiene

Orchestration, Planning, and Multi-Step Workflows

• Chaining actions, utilizing subagents, and decomposing tasks
• Comparing planning algorithms with heuristic orchestration
• Addressing failures, implementing retries, and executing compensating actions

Safety, Testing, and Observability

• Threat modeling, red-teaming, and sanitizing inputs and outputs
• Conducting unit, integration, and end-to-end testing for agents
• Implementing logging, metrics, tracing, and alerting for agent behavior

Deployment, Scaling, and MLOps for Agents

• Containerization, CI/CD pipelines, and effective rollout strategies
• Controlling costs, applying rate limiting, and optimizing resources
• Monitoring, governance, and operational playbooks

Summary and Future Directions