← ML_LAB

Visual References

Polished diagrams covering key concepts from the AWS Certified AI Practitioner (AIF-C01) curriculum. Each diagram is paired with a brief explanation of the underlying concept.

Certification Prep 03 diagrams
01

Confusion Matrix & Classification Metrics

A confusion matrix organizes a classifier's predictions against actual outcomes into four quadrants: True Positives, False Positives, True Negatives, and False Negatives. From these counts, we derive key metrics — accuracy measures overall correctness, precision captures how many predicted positives are truly positive, and recall captures how many actual positives were found. The diagram below uses a COVID-19 testing scenario to make these abstractions concrete: out of seven people tested, you can trace exactly how each metric is calculated.

Confusion matrix diagram showing COVID-19 testing example with TP, FP, TN, FN quadrants and derived accuracy, precision, and recall metrics
02

Simple RAG Pipeline

Retrieval-Augmented Generation (RAG) grounds an LLM's responses in your own data. This diagram traces the end-to-end workflow on AWS: documents in S3 are chunked and embedded using Amazon Titan, then stored as vectors in OpenSearch. At query time, the user's question is embedded and a similarity search retrieves the most relevant chunks. These are injected into the prompt context alongside the original question, and an LLM (such as Amazon Nova) generates a grounded response. The pipeline closes with Bedrock Evaluation checking faithfulness, answer relevance, and ROUGE scores.

End-to-end RAG pipeline diagram showing data ingestion from S3, embedding with Titan, vector storage in OpenSearch, similarity search, prompt augmentation, LLM generation, and Bedrock evaluation
03

Complete Lifecycle of Large Language Models

This diagram maps the full LLM lifecycle. It starts with pre-training on massive unstructured data (web crawls, books, articles) to produce a foundation model. From there, a key decision: use the model as-is, fine-tune it with labeled domain data for specialized tasks, or augment it with RAG to ground outputs in enterprise knowledge. Prompt engineering techniques — zero-shot, chain-of-thought — and parameters like temperature and top-k further shape model behavior. The cycle closes with evaluation: traditional metrics (ROUGE, BLEU, BERTScore) for measurable tasks and LLM-as-a-judge for subjective qualities like faithfulness and relevance.

Complete LLM lifecycle diagram showing pre-training, fine-tuning, RAG, prompt engineering, and evaluation with traditional metrics and LLM-as-a-judge
// About this playbook

Visual References

Diagrams covering confusion matrices, RAG pipelines and the complete LLM lifecycle, drawn for the AIF-C01 curriculum.

It covers confusion matrices, retrieval-augmented generation pipelines, and the full foundation-model lifecycle from pre-training to deployment. It is written for visual learners, and anyone who needs a diagram to put in a slide.

Every branch states the trade-off that decided it, so the recommendation you end on comes with the reasoning attached — something you can paste into a design note or defend in a review.

Exam prep Diagrams 10 min Certification Pass a certification

Related playbooks

Who made this

Built by Tarek Atwan — twenty years in data and AI, four books, four-time Pluralsight Elite instructor, Fortune 500 engagements across eight countries. Consulting through Ensemble Methods. Source on GitHub.