Written by Rahish · AI/ML Engineer
100% FREE
Master Production AI. One Readable Path at a Time.
Built like an engineering handbook, not a 40-hour video dump. Every lesson gives you the core mental model, copy-ready Python code, expected terminal output, “Watch out” production failure modes, and a hands-on capstone project.
CURRICULUM OVERVIEW
From Blank Terminal to Multi-Agent Systems
6 Tracks
Beginner to Flagship Agents
54 Lessons
~34 hours of deep practice
6 Projects
Real data, RAG & agent builds
$0 Forever
No paywalls or sign-up gates
01 / Recommended Roadmap
Python for AI
venv, types, JSONL, HTTP & data prep
LLM Fundamentals
Tokens, sampling, roles, cost & latency
Prompt Engineering
System prompts, schemas & eval loops
Building RAG Systems
Chunking, embeddings, hybrid search
GenAI & Agents
Tool calling, ReAct, HITL & evals
Azure AI Foundry
Deploy, ground, govern & ship on Azure
02 / Full Curriculum
Choose your engineering track.
Every track includes a sidebar lesson player, copy-ready code blocks, and a real project you can push to GitHub.
Python for AI Developers
Setup, virtual environments, type hints, dicts/JSONL, file paths, HTTP APIs with retries, and NumPy/pandas — the exact subset of Python you need before calling an LLM.
LLM Fundamentals
How language models actually generate tokens — tokenization, context windows, next-token probabilities, system/developer/user roles, temperature & sampling, and cost/latency budgets.
Prompt Engineering for Engineers
Treat prompts like versioned software specs — instruction hierarchy, few-shot selection, schema-constrained JSON outputs, prompt injection defenses, and automated prompt evals.
Building RAG Systems
End-to-end Retrieval-Augmented Generation — structure-aware chunking, dense embeddings, vector databases, hybrid BM25 + vector search, reranking, score floors, and inline citations.
Generative AI & Autonomous Agents
Go from single LLM calls to production multi-step agents — streaming, conversation memory, deterministic tool calling, ReAct loops, human-in-the-loop approval gates, and automated agent evaluation suites.
Flagship Capstone Project
Multi-Step Research Agent with Tool Guardrails
Build an agent that searches, verifies sources, respects step/token budgets, pauses for human approval on sensitive actions, and logs full traces.
Start Flagship TrackAzure AI Foundry (Microsoft Foundry)
Ship AI on Azure the enterprise way — resources and projects, keyless Entra ID auth and RBAC, the model catalog and deployment types, RAG with Azure AI Search, the Foundry Agent Service and tools, evaluation and red teaming, guardrails, tracing, and cost control.
Capstone Project
Grounded Customer-Support Agent on Foundry
Build an agent that answers from a knowledge base with citations, checks orders through function tools, escalates to humans, and ships behind an evaluation gate with full tracing.
Start Azure Track