AIAI study plan

Define core AI, ML, and deep learning terms in your own words and map how they relate to one another

Distinguish supervised, unsupervised, and reinforcement learning and identify which paradigm fits three sample problems

Describe the stages of the ML model lifecycle from problem framing through monitoring

Explain the bias-variance tradeoff and overfitting to a non-technical colleague in two minutes

Compare at least five common model families by use case, data needs, and interpretability

Select appropriate evaluation metrics for classification and regression scenarios and justify each choice

Explain how train/validation/test splits and cross-validation prevent leakage

Identify the causes of model drift and describe a monitoring approach for each

Audit a sample dataset for missing values, duplicates, outliers, and inconsistent formats

Apply at least three cleaning and transformation techniques and document the before/after effect

Detect and remediate class imbalance in a labeled dataset

Write data quality checks that would catch common pipeline failures

Explain how transformer architecture and attention enable large language models

Distinguish pretraining, fine-tuning, and retrieval-augmented generation by cost, control, and use case

Identify and mitigate at least four common LLM failure modes such as hallucination and prompt injection

Draft effective prompts that specify role, constraints, and output format

Identify sources of bias at each stage of the AI lifecycle and propose a mitigation for each

Apply a fairness metric to a scenario and explain the tradeoffs between definitions

Map privacy risks to controls such as anonymization, minimization, and access limits

Describe threat models for AI systems including data poisoning and model extraction

Outline the governance roles and documentation needed for an AI system

Evaluate five business use cases for AI feasibility using data readiness, value, and risk

Calculate a simple ROI estimate for one proposed AI initiative

Distinguish where AI adds value from where rules-based automation is better

Translate a business problem into AI requirements and success criteria

Compare deployment patterns such as batch, real-time API, and edge by latency and cost

Describe CI/CD and MLOps practices needed to ship and maintain a model

Design a monitoring plan covering performance, drift, and data quality

Identify rollback and incident-response steps for a failing model in production

Complete two timed practice exams and score each domain separately

Re-study your three weakest domains using targeted notes

Recite key definitions and frameworks for each domain from memory

Review every missed question and write the underlying concept it tested