Curriculum
- 11 Sections
- 2 Lessons
- 4 Weeks
Expand all sectionsCollapse all sections
- Module 1 — AI and LLM foundations Week 1 · FoundationsLLMs, transformers, tokens, embeddings and context windows System, user and tool messages Temperature, structured outputs and hallucinations Prompt engineering, JSON schemas and API fundamentals Model selection, latency, context limits and token costs1
- Module 2 — Python and LLM application developmentPython functions, classes, typing, async/await and exceptions Pydantic models and validation Environment variables and secret management LLM SDKs, REST clients and retries Logging, unit tests and configuration management1
- Module 3 — Agent fundamentals and tool callingWhat distinguishes an agent from a chatbot Observe, reason, select action, execute and observe again Function/tool calling and tool-result handling ReAct-style agent loops Tool schemas, error recovery, iteration limits and stop conditions0
- Module 4 — Agent orchestration with LangGraph- Graphs, nodes, edges and conditional routing - Shared state, reducers and checkpoints - Loops, branching and parallel execution - Durable workflows, retries and resumability - Human-in-the-loop approval and interrupt/resume patterns0
- Module 5 — RAG, memory and knowledge retrieval- Document parsing, chunking and metadata - Embeddings, vector search, hybrid search and reranking - Grounded answers, citations and access control - Conversation memory versus persistent knowledge - GraphRAG, knowledge graphs and entity relationships - Retrieval evaluation and handling missing evidence0
- Module 6 — Multi-agent systemsSupervisor and worker patterns Planner, researcher, executor and reviewer roles Sequential and parallel agent execution Shared state, message passing and delegation Agent handoffs, loops and termination When a multi-agent design is unnecessary0
- Module 7 — MCP and enterprise integrationsMCP clients, servers, tools, resources and prompts Tool discovery and schema validation Connecting internal APIs and enterprise data Authentication, authorization and least privilege Remote integrations, timeouts and untrusted tool output0
- Module 8 — Security, guardrails and human oversightPrompt injection and malicious documents Tool allowlists and permission boundaries PII protection and secrets management Approval gates for external or irreversible actions Tenant isolation, audit logs and data retention Threat modeling and security testing0
- Module 9 — Evaluation and observability- Unit, integration and end-to-end tests - Trace analysis of model calls and tool invocations - Retrieval quality and task-completion metrics - Hallucination, regression and adversarial testing - Token usage, latency, cost and failure monitoring - Offline evaluation datasets and production feedback0
- Module 10 — Production deployment- FastAPI and asynchronous request handling - Docker, configuration and CI/CD - AWS or Azure deployment patterns - Queues, background workers and rate limits - Databases, checkpoint persistence and scaling - Observability, incident response and cost controls0
- Module 11 — Enterprise agent designAgents versus deterministic workflows Event-driven systems and long-running jobs Idempotency, retries, compensating actions and approvals Multi-tenancy, auditability and SLAs Build-versus-buy and model selection Business ROI, operational costs and risk management0
Python and LLM basics
Next