AAI-M study plan

Define and contrast core AI terms (narrow vs. general AI, ML vs. deep learning, model vs. algorithm) in your own words

Classify ten real-world systems as rule-based, classical ML, or generative AI and justify each label

Explain the training-to-inference pipeline and identify where data, model, and human steps sit

Write plain-language definitions for at least 25 domain-1 terms without copying source text

Distinguish supervised, unsupervised, and reinforcement learning and match each to a use case

Describe the model lifecycle stages from problem framing through monitoring and explain what happens at each

Identify overfitting, underfitting, and data leakage from described scenarios

Explain how train/validation/test splits and cross-validation reduce evaluation error

Select appropriate metrics (accuracy, precision, recall, F1, RMSE) for stated business scenarios and justify the choice

Evaluate a described dataset for completeness, consistency, timeliness, and fitness for purpose

Distinguish data types (structured, unstructured, semi-structured) and describe handling implications

Propose concrete remediation steps for three named data-quality defects

Explain how transformer-based LLMs use tokens, embeddings, and attention at a conceptual level

Distinguish pretraining, fine-tuning, and prompting and state when each is appropriate

Identify hallucinations, context-window limits, and prompt-injection risks in given examples

Describe retrieval-augmented generation and when it beats fine-tuning

Map AI capabilities to business value categories such as cost reduction, revenue growth, and risk reduction

Build a value hypothesis for a use case stating baseline, expected impact, and how it would be measured

Distinguish feasible use cases from hype by applying data, cost, and risk filters

Identify stakeholders and success metrics for a proposed deployment

Define bias, fairness, transparency, and explainability and give an applied example of each

Identify sources of bias across data collection, labeling, and deployment stages

Apply at least two fairness metrics or mitigation strategies to a described scenario

Explain human-in-the-loop and escalation design for high-stakes decisions

Classify data in a scenario as personal, sensitive, or non-personal and state handling obligations

Explain key privacy principles (minimization, purpose limitation, retention) and apply them to an AI workflow

Describe governance artifacts such as model inventories, risk assessments, and audit trails and their purpose

Outline an incident-response path for a data or model failure

Sequence AI project delivery phases and identify the key decision gate at each

Describe monitoring, drift detection, and retraining triggers for a deployed model

Revisit earlier weak areas and close gaps identified in practice assessments

Complete timed practice sets across all eight domains and review every missed item