Python AI Engineering Practice

A structured 65-question coding curriculum for Full-Stack AI Engineers — built to develop real muscle memory writing Python from scratch, without autocomplete.

Covers everything from Python fundamentals to production-grade RAG pipelines, LangGraph agents, vector databases, and Docker.

---

Structure

Level	Folder	Topics	Questions
1	root	Syntax, I/O, conditionals, loops	Q1–Q5
2	root	Lists, dicts, sets, collections	Q6–Q10
3	root	Functions, JSON, FastAPI basics	Q11–Q15
4	level_2/	Async, FastAPI advanced, RAG, LangChain, LangGraph	Q16–Q20
5	level_3_oop/	Classes, inheritance, decorators, ABC, context managers	Q21–Q30
6	level_4_power_features/	Comprehensions, generators, *args, lambda, type hints	Q31–Q40
7	level_5_production/	File I/O, logging, env vars, requests, Pydantic, SQLAlchemy	Q41–Q50
8	level_6_ai_engineering/	RAG deep dive, LangGraph advanced, vector DBs, streaming, Docker	Q51–Q65

---

Topics Covered

Python Fundamentals

• Conditionals, loops, exception handling, user input
• String manipulation, list/dict/set operations
• Functions, return values, variable scope

OOP

• Classes, __init__, dunder methods (__str__, __repr__, __len__, __contains__)
• Inheritance, polymorphism, abstract base classes
• @property, @classmethod, @staticmethod, @dataclass
• Context managers (__enter__/__exit__ and @contextmanager)

Python Power Features

• List, dict, and set comprehensions
• Decorators (@timer, @log_calls)
• Generators with yield
• *args / **kwargs, lambda, zip(), enumerate()
• Type hints and annotations

Production Patterns

• File I/O, JSON read/write, environment variables (.env)
• Structured logging with file + console handlers
• HTTP requests (requests, aiohttp, async concurrent fetching)
• Pydantic models with validation
• SQLAlchemy ORM: model, session, CRUD
• pytest: unit tests, error assertions, parametrize

FastAPI

• Basic POST endpoint with Pydantic request/response models
• Dependency injection, async endpoints
• Full CRUD with in-memory store
• StreamingResponse + Server-Sent Events (SSE)

AI Engineering

• RAG pipeline: chunking → embedding → ChromaDB → retrieval → prompt → LLM
• LangChain: prompt templates, LCEL pipe operator (|), StrOutputParser
• LangGraph: TypedDict state, node functions, StateGraph, conditional edges
• Multi-agent pattern: researcher → writer → critic loop
• Text chunking: fixed size with overlap, by sentence, by paragraph
• Token counting with tiktoken, context window management
• Nested JSON parsing from OpenAI-style API responses
• Exponential backoff retry for LLM API calls
• Prompt hydration with safe variable injection

Databases & Infrastructure

• MongoDB: insert, find, update, delete
• Redis: LLM response caching with TTL and query hash keys
• Qdrant: create collection, ingest vectors, similarity search, metadata filtering
• Docker: Dockerfile, docker-compose.yml (FastAPI + Redis + Qdrant)

---

Running the Code

# Setup
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Run any standalone script
python q01_odd_even.py

# Run FastAPI apps
uvicorn q15_fastapi_endpoint:app --reload
uvicorn level_2/q17_fastapi_advanced:app --reload

# Run tests
pytest level_6_ai_engineering/q64_pytest_basics.py -v

---

Stack

Python 3.13 · FastAPI · LangChain · LangGraph · ChromaDB · Qdrant · Redis · MongoDB · Pydantic · SQLAlchemy · tiktoken · aiohttp · pytest · Docker