Building real applications on large language models, in 30 parts, vendor neutral and in Python. This series starts with calling a model properly, then builds in order: retrieval and RAG, tool calling and agents, evaluation and safety, then the production concerns of cost, latency, deployment and provider choice. A single internal support assistant project runs through the whole series, so each part builds on the one before. It assumes the GenAI Series for concepts like tokens, attention and what RAG is, and the Data Science Series for Python, rather than repeating them.

1. 01What an AI Engineer Actually Does, vs ML Engineer, Data Scientist and Backend Developer
2. 02Calling a Model From Python: SDKs, Messages and Parameters That Matter
3. 03Prompt Patterns That Survive Contact With Real Users
4. 04Structured Output: JSON Mode, Schemas and Validation That Does Not Fail Silently
5. 05Token Budgets, Context Limits and Cost Per Request
6. 06Errors, Retries, Rate Limits and Timeouts

1. 07Why Retrieval Beats Fine Tuning for Most Business Problems
2. 08Document Ingestion: Parsing, Cleaning and Formats That Break
3. 09Chunking Strategies and How to Choose One
4. 10Embedding Models: Choosing, Benchmarking and What They Cost
5. 11Vector Stores in Practice: pgvector, Chroma and Qdrant Compared
6. 12Hybrid Search and Reranking for RAG Retrieval
7. 13Retrieval Evaluation, Separated From Answer Evaluation

1. 14Tool Calling From First Principles
2. 15Agent Loops and Planning, and When an Agent Is the Wrong Answer
3. 16Model Context Protocol and Standardising Tool Access
4. 17Multi Step Workflows, State and Memory
5. 18Multi Agent Patterns and Their Coordination Cost
6. 19Human in the Loop and Approval Gates

1. 20Building an Eval Set Before You Build the Feature
2. 21Automated Evaluation: LLM as Judge, and Where It Lies to You
3. 22Regression Testing Prompts and Model Upgrades
4. 23Guardrails: Input Filtering, Output Validation and PII
5. 24Prompt Injection and the Security Model of an LLM Application
6. 25Observability: Tracing and Debugging a Non Deterministic System

1. 26Caching, Batching and Latency Engineering
2. 27Cost Control and Model Routing
3. 28Deployment, Versioning and Rollback for Prompts and Models
4. 29Choosing and Switching Providers Without a Rewrite
5. 30Path to AI Engineer, and What to Learn Next