AI Engineering Course

A project-based course to go from zero to $200k+ AI Engineer, working remotely from Brazil.

Goal: Build real things every module. No toy examples. Everything goes in a portfolio.

---

Roadmap

#	Module	Key Skill	Project	Status
01	LLM API Fundamentals	First real API calls, system prompts, structured outputs	CLI chat app	✅
02	Prompt Engineering	Chain-of-thought, few-shot, formatting, reliability	Prompt testing harness	✅
03	Embeddings & Semantic Search	Vectors, cosine similarity, pgvector	Semantic search over docs	✅
04	RAG — Retrieval Augmented Generation	Chunking, retrieval, reranking	RAG Q&A over a PDF corpus	✅
05	AI Agents & Tool Use	Function calling, tool loops, ReAct	Research agent	✅
06	Evals — Measuring LLM Quality	LLM-as-judge, regression testing, benchmarks	Eval suite for your RAG	✅
07	Streaming & Production Patterns	Streaming, async, error handling, caching	Production-ready API	✅
08	Fine-tuning	LoRA, QLoRA, datasets, PEFT	Fine-tune a model on custom data	✅
09	Multi-agent Systems	Orchestration, parallelism, handoffs	Multi-agent research pipeline
10	Observability & LLMOps	Tracing, cost tracking, latency, monitoring	Instrumented production app
11	Multimodal — Vision & Audio	Image inputs, audio transcription, OCR	Document intelligence app
12	System Design for AI	Architecture, trade-offs, scalability	Design doc for a real system
13	Capstone	Everything	Ship a complete AI product

---

How to use this course

1. Do modules in order. Each builds on the previous.
2. Build the project before reading the solution. Struggle is part of it.
3. Commit everything. This repo is your portfolio.
4. Ship each project. Deploy it, write a short post, put it on LinkedIn.

---

Stack

• Language: Python 3.12+
• Primary API: Anthropic (Claude) — also covers OpenAI where needed
• Vector DB: pgvector (local via Docker), Pinecone for production
• Framework: Mostly from scratch, then LangChain/LangGraph where it earns its place
• Deploy: FastAPI + Railway/Render (free tier friendly)
• Eval: Braintrust or custom harness
• Observability: Langfuse (self-hosted) or LangSmith

---

Setup

# Python 3.12+
python -m venv .venv
source .venv/bin/activate

pip install anthropic openai python-dotenv fastapi uvicorn

Create .env:

ANTHROPIC_API_KEY=sk-ant-...
OPENAI_API_KEY=sk-...

---

Career strategy (Brazil → $200k remote)

The target is US-based companies that hire internationally. That means:

• GitHub portfolio matters more than a degree. Every project here is a public artifact.
• Target companies: AI-native startups (Series A–C), not FAANG. They pay well and hire globally.
• Job boards: Wellfound (AngelList), Otta, Greenhouse, direct outreach on LinkedIn.
• Rate: As a contractor (PJ), aim for $80–120/hr. As a full-time remote employee, $150–200k base + equity.
• Stack to emphasize on resume: RAG, agents, evals, LangChain/LangGraph, Claude/OpenAI APIs, production deployment.

After Module 6, you are already hirable for a mid-level role. After Module 10, senior.

---