Coders {cookies} · Open-source skills, free for everyone

Overview

This is the master AI track. It merges your two source roadmaps into one sequence and keeps every topic from both. It assumes intermediate Python by month two. Each week ships a small project so the learning compounds. The path runs foundations, then data and embeddings, then generative AI and prompting, then RAG and frameworks, then machine learning and fine tuning, and finally agents, evaluation, production, and a capstone.

Your original files are preserved in the workspace root as the detailed source. This file is the structured, app ready version.

Month 1: Python, Git, and engineering foundations

Week 1: Python fundamentals

0 of 5

Variables, data types, conditionals, loops, comprehensions
How Python executes code: interpreter and bytecode
Debugging with pdb, reading tracebacks, the logging module
Virtual environments: venv, conda, uv
Type hints and mypy

Tools and libraries

Python 3.12 or newer, uv, ruff, mypy

Build

A CLI calculator and a number guessing game with uv and type hints

Week 2: Functions, data structures, and pythonic code

0 of 5

Functions, closures, decorators, args and kwargs
Lists, dicts, sets, tuples and when to use each
Generators and iterators for streaming
dataclasses, namedtuple, TypedDict
collections: defaultdict, Counter, deque

Tools and libraries

pydantic v2, rich, typer

Build

A marks analyzer with pydantic models and a typer to do CLI

Week 3: OOP, design patterns, and files

0 of 5

Classes, inheritance, dunder methods
SOLID principles
Patterns: factory, singleton, strategy, observer
File IO: CSV, JSON, YAML, TOML
pathlib for cross platform paths

Tools and libraries

loguru, python dotenv, orjson

Build

A library management system with OOP, loguru, and a YAML config

Week 4: Git, APIs, and testing

0 of 5

Git: commit, branch, merge, rebase, cherry pick, stash
Pull requests, code review, and CI basics
REST APIs with httpx
pytest: unit tests, fixtures, mocking
Pre commit hooks

Tools and libraries

httpx, pytest, pytest asyncio, pre-commit, GitHub Actions

Build

A weather CLI with pytest tests and a CI pipeline

Month 2: Data, text, and embeddings

Week 5: Data handling

0 of 5

pandas: load, filter, groupby, merge, pivot
polars and lazy evaluation
EDA and profiling
Visualization with matplotlib, seaborn, plotly
Working with Parquet

Tools and libraries

pandas, polars, plotly, pyarrow

Build

Analyze a real dataset with pandas and polars and export to Parquet

Week 6: Data cleaning, features, and pipelines

0 of 5

Missing value imputation
Feature engineering: encoding, scaling, binning
scikit-learn preprocessing pipelines
Schema validation with pandera
Data version control basics

Tools and libraries

pandera, dvc, category encoders

Build

Clean a messy dataset with a pandera schema and DVC tracking

Week 7: Text processing and NLP foundations

0 of 4

Tokenization, stopwords, lemmatization, stemming
TF IDF and BM25 for keyword search
Named entity recognition with spaCy
Regex for text cleaning

Tools and libraries

spaCy, rank bm25, nltk

Build

A text pipeline with tokenization, entity extraction, and a BM25 index

Week 8: Embeddings and semantic search

0 of 5

Embeddings, vector spaces, cosine similarity
Sentence transformers and SBERT
OpenAI embedding models
Hybrid search: BM25 plus dense with reciprocal rank fusion
Choosing a model from the MTEB leaderboard

Tools and libraries

sentence transformers, faiss, the OpenAI SDK

Build

Semantic search over your own notes with FAISS and SBERT

Month 3: Generative AI and prompt engineering

Week 9: LLM fundamentals and architecture

0 of 5

Transformer architecture: attention, positional encoding
Tokenization: BPE, SentencePiece, tiktoken
Context windows, KV cache, inference throughput
LLM families and trade offs
Quantization: INT4, INT8, GPTQ, AWQ, GGUF

Tools and libraries

tiktoken, Ollama, LM Studio

Build

A chatbot using a hosted API and a local model, comparing outputs

Week 10: Prompt engineering

0 of 6

Zero shot, few shot, chain of thought, self consistency
ReAct prompting
Tree of thoughts and least to most
Structured outputs: JSON mode and tool schemas
Prompt injection and jailbreak defense
Programmatic prompt optimization with DSPy

Tools and libraries

DSPy, instructor, promptflow

Build

A document summarizer and JSON extractor using instructor

Week 11: Multi provider GenAI APIs

0 of 5

OpenAI, Anthropic, Gemini, Mistral, and Groq APIs
Streaming and server sent events
Token counting, rate limits, retries with backoff
Cost optimization: caching and batching
A unified API across providers with litellm

Tools and libraries

litellm, the OpenAI and Anthropic SDKs, tenacity

Build

A multi model API wrapper with cost tracking

Week 12: Function calling, tool use, and structured outputs

0 of 5

Tools and libraries

instructor, pydantic v2, structured outputs

Build

A PDF data extractor with typed pydantic models

Month 4: RAG, vector stores, and frameworks

Week 13: RAG fundamentals and chunking

0 of 5

Why RAG: knowledge cutoffs and hallucination
Ingestion versus retrieval pipelines
Chunking: fixed, recursive, semantic, document aware
Metadata filtering and tagging
HyDE

Tools and libraries

unstructured, docling, pdfplumber

Build

A robust document chunking pipeline for PDF, DOCX, and HTML

Week 14: Vector databases

0 of 5

Indexing: flat, IVF, HNSW
Distance metrics: cosine, dot product, euclidean
Filtered and hybrid search
Namespaces and multi tenancy
Production: indexing speed, memory, scale

Tools and libraries

Qdrant, Pinecone, Chroma, pgvector, Weaviate

Build

Vector search over a thousand documents deployed to Qdrant

Week 15: Advanced RAG

0 of 6

Naive to advanced to modular RAG
Query transformations: HyDE, multi query, step back
Reranking with cross encoders
The lost in the middle problem
RAG evaluation with RAGAS
Agentic RAG patterns

Tools and libraries

ragas, Cohere Rerank, flashrank

Build

A document question and answer bot with query expansion, reranking, and RAGAS evaluation

Week 16: Frameworks and observability

0 of 5

LangChain expression language, retrievers, memory
LlamaIndex: nodes, query engines, routers
Framework versus from scratch trade offs
Tracing with LangSmith and Langfuse
OpenTelemetry for AI apps

Tools and libraries

langchain, llama-index, langfuse

Build

A production style RAG with full tracing through Langfuse

Month 5: Machine learning for AI engineers

Week 17: ML fundamentals and math

0 of 6

Supervised, unsupervised, self supervised, RLHF
Loss functions: MSE, cross entropy, contrastive
Optimization: Adam, learning rate schedules
Bias and variance trade off
Linear algebra: matrices, eigenvectors, SVD
Calculus: partial derivatives, chain rule

Build

Linear regression from scratch with NumPy

Week 18: Core ML models

0 of 5

Linear and logistic regression, decision trees, random forests
Gradient boosting: XGBoost, LightGBM
K means and PCA
Hyperparameter tuning with Optuna
Explainability with SHAP

Tools and libraries

scikit-learn, xgboost, optuna, shap

Build

House price prediction with XGBoost and a SHAP report

Week 19: Evaluation and experiment tracking

0 of 5

Classification metrics: accuracy, F1, AUC ROC, PR curves
Cross validation and splits
Diagnosing overfitting and underfitting
MLflow tracking and a model registry
Weights and Biases

Tools and libraries

mlflow, wandb, evidently

Build

An experiment with MLflow tracking, a registry, and an eval dashboard

Week 20: Fine tuning

0 of 6

Tools and libraries

transformers, peft, trl, unsloth

Build

Fine tune a small Llama model with QLoRA and deploy it to a cloud GPU

Month 6: Agentic systems, production, and capstone

Week 21: Agentic AI systems

0 of 6

Architectures: ReAct, plan and execute, reflexion
Multi agent: supervisor, swarm, hierarchical
Tool design: idempotency, error handling, retries
Memory: working, episodic, semantic, procedural
Human in the loop patterns
The model context protocol

Tools and libraries

langgraph, crewai, smolagents, the MCP SDK

Build

An agent that automates a real workflow such as research, write, and email a report

Week 22: Evaluation, evals, and cost

0 of 6

LLM as judge evaluation
Golden datasets: create, maintain, version
deepeval, promptfoo, ragas
Metrics: faithfulness, relevance, context precision
Cost optimization: caching, batching, model routing
Prompt compression

Tools and libraries

deepeval, promptfoo, LLMLingua

Build

An evaluation harness with a cost per query report

Week 23: Production readiness and MLOps

0 of 7

Structured logging with correlation ids
Distributed tracing with OpenTelemetry
Prompt versioning and management
Docker and Compose for AI apps
Cloud deployment on AWS, GCP, or Azure
Responsible AI: PII detection, moderation, injection defense
Rate limiting and guardrails

Tools and libraries

Docker, guardrails ai, presidio, Railway or Render

Build

A deployed AI app with Langfuse observability and PII filtering

Week 24: Capstone and portfolio

0 of 2

Pick one: compliance checker, personal knowledge base, code review agent, support bot, or research assistant
Deliverables: a repo with Docker and a README, an architecture diagram, a one page write up, and a live demo

Build

A full AI product end to end with RAG or agents, a backend, observability, and a deployment

Deep dive supplements (from the twelve week source)

These topics from your second source file are folded in where relevant above, and listed here so none are missed:

Attention internals: Q, K, V matrices, multi head attention, feed forward sublayer, layer norm, residual connections

Positional encodings: sinusoidal, rotary, ALiBi

Encoder only, decoder only, and encoder decoder architectures

Attention visualization with BertViz

FastAPI backend for AI: routers, dependency injection, middleware, WebSockets, background tasks

Async Python and streaming LLM responses

PostgreSQL with asyncpg and SQLAlchemy, Redis caching, Supabase auth and pgvector

LLM serving: Ollama, vLLM, litellm proxy

CI and CD, Docker Compose multi service stacks, GCP Cloud Run, Nginx, Prometheus, Grafana

Advanced fine tuning: RLHF, DPO, model merging with mergekit, GGUF and AWQ quantization, LM Evaluation Harness

AI paper reading methodology and key papers

Resource master reference

Books

Designing Machine Learning Systems by Chip Huyen

Hands On Machine Learning by Aurelien Geron

Fluent Python by Luciano Ramalho

Free courses

fast.ai Practical Deep Learning

the Hugging Face NLP course

DeepLearning.AI short courses

Key papers

Attention Is All You Need

the RAG paper

ReAct

LoRA

Constitutional AI

Lost in the Middle

Tools master list

uv, ruff, mypy, pytest, loguru, httpx, litellm, instructor, dspy, langgraph, deepeval, langfuse, Qdrant, docling, sentence transformers, Cohere Rerank, ragas, scikit-learn, xgboost, transformers, peft, trl, unsloth, wandb, mlflow

Interview focus

Implement attention in NumPy

Build a simple BPE tokenizer

Cosine similarity search with FAISS

A streaming API wrapper with retry logic

Design a RAG search system for ten million users

RAG versus fine tuning and when to use which

Preventing hallucinations in production