Curriculum Overview: Source Citation and Documenting Data Origins

Welcome to the curriculum overview for Source Citation and Documenting Data Origins, a critical governance pillar in the AWS Certified AI Practitioner (AIF-C01) exam. This curriculum focuses on data lineage, cataloging, and comprehensive model documentation tools like Amazon SageMaker Model Cards.

Prerequisites

Before diving into this curriculum, learners should have a solid foundation in the following areas:

Machine Learning Lifecycle: Familiarity with the stages of ML development, including data collection, preprocessing, feature engineering, model training, and deployment.
Generative AI Basics: Understanding the difference between training data, fine-tuning data, and user data.
Basic Cloud Concepts: Awareness of AWS services, particularly the Amazon SageMaker ecosystem (though deep technical expertise is not required upfront).
Data Governance Fundamentals: A conceptual understanding of why organizations care about data privacy, compliance, and security.

[!NOTE] If you are new to the ML lifecycle, consider reviewing the fundamentals of data preprocessing (e.g., cleaning, curating, and transforming) before starting this module.

Module Breakdown

This curriculum is structured to take you from foundational concepts of data provenance to practical implementation using AWS tools.

Module	Title	Difficulty	Core Focus
Module 1	Foundations of Data & Model Lineage	Beginner	Defining lineage, provenance, and their role in compliance.
Module 2	Source Citation vs. Data Origins	Intermediate	Differentiating between licensing/citation and deep transformation documentation.
Module 3	Data Cataloging Strategies	Intermediate	Building a systemic "library" for datasets and AI resources.
Module 4	Amazon SageMaker Model Cards	Advanced	Centralizing documentation, intended uses, and risk metrics.
Module 5	Model Registry & Version Control	Advanced	Managing model iterations and deployment approvals in SageMaker.

Loading Diagram...

Learning Objectives per Module

Module 1: Foundations of Data & Model Lineage

Define data and model lineage as the complete history of data origins and applied transformations.
Explain how data lineage supports governance, security, and compliance by surfacing hidden biases and quality issues.

Module 2: Source Citation vs. Data Origins

Source Citation: Properly acknowledge training data sources and document any licenses, permissions, or terms of use.
Documenting Data Origins: Record the granular details of how data was collected, curated, cleaned, and preprocessed.

Module 3: Data Cataloging Strategies

Organize datasets, models, and resources systematically.
Understand how a well-kept catalog improves internal management, auditing, and stakeholder communication.

Module 4: Amazon SageMaker Model Cards

Create standardized documentation for machine learning models.
Detail intended uses (and unintended uses), known biases, and performance benchmarks.
Assign risk ratings (Unknown, Low, Medium, High) to assess deployment impacts.

Module 5: Model Registry & Version Control

Utilize Amazon SageMaker Model Registry to catalog production-ready models.
Implement custom lifecycle stages (e.g., development, testing, production) and metadata association.

Loading Diagram...

Success Metrics

How will you know you have mastered this curriculum? You will be evaluated against the following milestones:

Data Provenance Mapping: Successfully trace and document the lifecycle of a mock dataset from its raw origin to its finalized training state, including all preprocessing steps.
Citation Compliance: Accurately audit a sample dataset to identify and record required licenses, permissions, and intellectual property constraints.
Model Card Creation: Draft a complete Amazon SageMaker Model Card for a Generative AI use-case. This includes explicitly defining the model's intended use cases, risk rating, and recording standard performance metrics.

For example, you must be able to document quantitative metrics (like Accuracy, BLEU, or $F_1$ scores) clearly in the Model Card's evaluation section:

$F_1 = 2 \times \frac{Precision \times Recall}{Precision + Recall}$

Registry Navigation: Demonstrate the ability to move a model through custom lifecycle stages (Development $\rightarrow$ Testing $\rightarrow$ Production) within the SageMaker Model Registry.

Real-World Application

In the real world, AI systems are only as trustworthy as the data they are built upon. Documenting data origins and implementing strong source citation is not just an academic exercise—it is a legal and operational imperative.

Mitigating Legal Risk: If a Generative AI model inadvertently violates copyright because it was trained on unlicensed data, the organization could face severe intellectual property infringement claims. Proper source citation prevents this.
Preventing Bias and Harm: By tracking data lineage and documenting data origins, ML engineers can surface underrepresented groups or toxic data early. This "defense in depth" protects end-users from biased outputs.
Ensuring Explainability: When an AI model makes a high-stakes decision (e.g., denying a loan application), regulators and auditors require explainability. Model Cards provide the single source of truth to prove the model was used within its intended scope and risk rating.
Career Readiness: Professionals who master these governance tools become invaluable assets. As an AI Governance Specialist, ML Engineer, or Cloud Compliance Officer, you will routinely rely on AWS services like SageMaker Model Cards and Data Cataloging to build trustworthy, enterprise-grade AI systems.

▶Click to expand: AWS Services for Governance Overview

While SageMaker handles model-level documentation, broader AWS governance is enforced via:

AWS Organizations and SCPs: Restricts use of certain services or organizational units.
AWS Config: Tracks configuration changes and compliance.
AWS CloudTrail: Logs API activity to enable audits and track accountability.