AIT study plan

Define and distinguish AI, machine learning, deep learning, and generative AI with one original example each

Classify AI systems by capability and by functionality using your own taxonomy table

Explain the standard ML workflow from problem framing through deployment and monitoring

Distinguish supervised, unsupervised, self-supervised, and reinforcement learning with a use case for each

Compute and interpret the bias-variance tradeoff for a described scenario

Select appropriate train/validation/test splits and cross-validation strategies for small and imbalanced datasets

Diagnose overfitting and underfitting from described training-curve behavior

Choose and justify evaluation metrics for classification, regression, and ranking tasks

Identify common data-quality defects and describe a remediation for each

Describe how feature engineering and dimensionality reduction affect model performance and interpretability

Trace forward and backward propagation through a small feedforward network by hand

Explain the role of activation functions, loss functions, and optimizers in training

Contrast CNNs, RNNs, and transformers by inductive bias and typical use case

Diagnose vanishing and exploding gradients and propose at least three mitigations

Explain regularization techniques for neural networks including dropout, weight decay, and early stopping

Describe transfer learning and when to fine-tune versus freeze layers

Interpret attention mechanisms and positional encoding in transformer models

Explain tokenization, embeddings, and contextual representations in NLP pipelines

Distinguish pretraining, fine-tuning, and prompting as adaptation strategies

Describe how large language models generate text and what causes hallucination

Compare retrieval-augmented generation with pure prompting for a knowledge-intensive task

Identify NLP evaluation approaches and their limitations

Apply fairness definitions and detect bias at data, model, and deployment stages

Map a described AI system to privacy, security, and accountability risks

Explain transparency, explainability, and human-oversight requirements for high-risk use cases

Outline a governance process covering documentation, audit, and incident response

Relate regulatory and risk-management frameworks to concrete engineering controls

Select deployment patterns such as batch, real-time, and edge, with tradeoff justification

Describe monitoring, drift detection, and model-retraining triggers in production

Explain CI/CD, versioning, and reproducibility practices for ML systems

Identify cost, latency, and scalability constraints for an AI workload

Integrate ethics review checkpoints into a deployment pipeline

Synthesize all domains into a single integrated review outline

Retake the Week 4 diagnostic and compare error patterns

Complete two timed full-length practice sets under exam conditions

Review and resolve every missed item in a consolidated error log