Static reference · Inline SVG figures

Machine learning & AI theory — illustrated

Conceptual notes for data engineering, classical machine learning, neural networks, production AI, RAG vs fine-tuning, and operations: MLOps, LLMOps, and foundational DevOps. Each page is self-contained with diagrams you can screenshot or print.

Use the cards below in any order. The RAG vs fine-tuning page is especially useful when choosing how to inject domain knowledge into language models.

Hub

Overview and navigation — you are here.

ml_theory_hub.html

Data engineering

Pipelines, warehouses, medallion layers, batch vs stream.

→ theory_data_engineering.html

Classical ML

Bias–variance, validation, metrics, ensembles — theory + visuals.

→ theory_classical_ml.html

Deep learning & NN

Neurons, layers, activations, training loop, inductive biases.

→ theory_deep_learning.html

Applied AI & AI engineering

From prototype to production: roles, MLOps loop, evaluation.

→ theory_applied_ai_engineering.html

RAG vs fine-tuning

When to retrieve vs when to update weights — decision view.

→ theory_rag_vs_finetune.html

MLOps

ML lifecycle: registry, training pipelines, deploy, monitor, retrain.

→ theory_mlops.html

LLMOps

Operating LLMs: prompts, RAG, eval harnesses, cost, safety.

→ theory_llmops.html

DevOps

CI/CD, IaC, observability — the base for MLOps and LLMOps.

→ theory_devops.html

How these relate

Data engineering moves and shapes data; classical ML often runs on tabular features in a warehouse; deep learning learns representations from raw signals; AI engineering ships and monitors models; RAG and fine-tuning specialize LLMs. DevOps enables reliable delivery; MLOps and LLMOps add ML- and LLM-specific automation on top.