Use the small amount of math that appears everywhere in AI: functions, vectors, matrices, distance, rates of change, and simple optimization. The goal is practical reading fluency, not formal proof.

Handle uncertainty with probability, distributions, Bayes’ rule, sampling, averages, variance, and confidence. These ideas support classification, prediction, decision-making, and model evaluation.

See how AI moved from early logic programs and expert systems to machine learning, deep learning, foundation models, and today’s AI products. This history explains why current practice mixes old ideas, big data, specialized chips, and new safety concerns.

Solve problems by trying possible paths, scoring them, and choosing better routes. You will use search ideas behind games, routing, puzzle solvers, scheduling, and many planning systems.

Represent facts, rules, relationships, and constraints in ways a computer can reason over. This chapter covers logic, inference, ontologies, and where rule-based AI still matters.

Model goals, states, actions, costs, and consequences so an AI system can choose a sequence of steps. You will connect planning to robots, logistics, game characters, and automated workflows.

Use probability to represent partial knowledge and competing explanations. You will work with Bayesian networks, Markov models, hidden states, and decision-making under risk.

Train models from labeled examples and use them to predict categories or numbers. You will practice the core loop of splitting data, fitting a model, testing it, and improving it.

Judge models with accuracy, precision, recall, F1, ROC curves, calibration, error analysis, and baselines. You will learn how to spot overfitting and avoid fooling yourself with a lucky test result.

Shape data so models can use it well through cleaning, encoding, scaling, missing-value handling, and feature creation. This chapter shows why careful data work often matters more than a fancier algorithm.

Use decision trees, random forests, gradient boosting, linear models, support vector machines, and nearest-neighbor methods. You will know when simple classical methods beat heavier approaches.

Find structure without labels using clustering, dimensionality reduction, anomaly detection, and topic-style grouping. These tools help with discovery, compression, monitoring, and messy datasets where answers are not pre-marked.

Build the mental model of a neural network: layers, weights, activations, loss functions, backpropagation, and gradient descent. You will train small networks and see why depth changes what models can represent.

Improve neural networks with regularization, normalization, better optimizers, data augmentation, transfer learning, and careful hyperparameter tuning. This chapter turns fragile demos into models that train reliably.

Use convolutional networks and modern vision methods for classification, detection, segmentation, and visual search. You will also learn common image data problems such as imbalance, lighting, labeling noise, and privacy.

Represent language with tokens, embeddings, sequence models, parsing, classification, extraction, and generation. This chapter builds the path from older NLP methods to the models used in current language tools.

Handle sound, speech-to-text, text-to-speech, speaker features, noise, and audio event detection. You will see how AI systems turn waveforms into useful signals and human-facing interfaces.

Model users, items, preferences, rankings, similarity, feedback loops, and cold-start problems. You will build the ideas behind recommendation feeds, search ranking, playlists, and product suggestions.

Work with data that changes over time using forecasting, seasonality, sequence features, recurrent models, and temporal validation. This chapter covers demand prediction, sensors, finance-like signals, and operations data.

Separate correlation from cause using experiments, counterfactual thinking, confounders, causal graphs, and treatment effects. You will see how causal AI supports policy, medicine, product decisions, and safer automation.

Train agents that learn by trying actions and receiving rewards. You will work with value functions, policies, exploration, simulation, and the tradeoffs that make reinforcement learning powerful but difficult.

Connect perception, planning, control, sensors, motors, and safety constraints. This chapter shows how AI leaves the screen in robots, drones, vehicles, warehouses, and assistive devices.

Use graphs to represent networks of people, products, documents, molecules, roads, and facts. You will work with graph search, graph embeddings, knowledge graphs, and graph neural networks.

Design the data pipeline behind AI work: collection, labeling, storage, versioning, lineage, quality checks, and dataset documentation. You will see how strong AI projects depend on strong data operations.

Use GPUs, TPUs, batching, distributed training, mixed precision, and memory-aware model design. This chapter explains the hardware and systems choices that made modern large-scale AI possible.

Use attention, positional information, encoders, decoders, and self-supervised training. Transformers get their own chapter because they changed language, vision, audio, biology, search, and generation.

Work with large pretrained models that can adapt to many tasks through prompting, examples, and context. You will cover model families, pretraining, scaling laws, capability limits, and why foundation models changed AI practice.

Write prompts that give models the right role, task, context, examples, format, and checks. You will also handle prompt injection, brittle instructions, evaluation, and when prompting is not enough.

Connect language models to search indexes, document stores, databases, and citations so answers can use fresh or private information. You will design chunking, embeddings, retrieval, reranking, grounding, and source-checking.

Adapt pretrained models with fine-tuning, instruction tuning, parameter-efficient methods such as LoRA, and preference tuning. This chapter covers when adaptation helps, what data it needs, and how to avoid damaging a capable model.

Create images, audio, video, and structured outputs with diffusion models and related generative methods. You will cover denoising, conditioning, guidance, editing, evaluation, rights issues, and misuse risks.

Combine text, images, audio, video, tables, and actions in one system. This chapter covers multimodal embeddings, vision-language models, document AI, visual question answering, and richer human-computer interaction.

Let models call tools, write intermediate steps, use memory, follow plans, and coordinate subtasks. You will learn where agentic systems are useful, where they fail, and how to add guardrails and human review.

Use AI to write, review, test, explain, and refactor code while keeping human responsibility for correctness and security. This chapter covers coding assistants, generated tests, repository context, and safe development habits.

Move a trained model into a real application with APIs, batch jobs, streaming, edge devices, latency targets, cost limits, and fallback behavior. You will compare deployment patterns and know what changes outside a notebook.

Track data drift, model drift, errors, latency, cost, bias, and user impact after launch. This chapter covers logging, observability, retraining triggers, incident response, and rollback plans.

Test AI systems beyond normal accuracy using red teaming, adversarial examples, prompt injection tests, data leakage checks, jailbreak attempts, and abuse cases. You will learn how attackers and careless users can break otherwise impressive systems.

Protect personal data and sensitive information with minimization, anonymization limits, differential privacy, federated learning, access controls, and secure data handling. This chapter connects privacy choices to model quality and legal risk.

Identify unfair outcomes, biased data, harmful feedback loops, accessibility issues, and unequal error rates. You will practice documenting tradeoffs, choosing fairness metrics, and designing human review where mistakes matter.

Work with model cards, dataset sheets, audit trails, risk ratings, human oversight, copyright questions, and AI laws such as the EU AI Act. This chapter turns ethics and compliance into concrete project duties.

Follow one complete project from a real request through problem framing, data checks, model choice, evaluation, deployment, monitoring, and stakeholder handoff. You will see how technical decisions, risk, cost, and user needs fit together.

Choose the right AI approach for a new problem by comparing rules, search, classical ML, deep learning, foundation models, retrieval, fine-tuning, and human-in-the-loop workflows. This chapter builds the judgment to avoid overbuilding or using AI where it does not belong.

Plan your next steps through roles such as AI engineer, machine learning engineer, data scientist, research engineer, product builder, safety specialist, and domain expert. You will map portfolio projects, open-source contribution paths, certifications, communities, papers, benchmarks, and habits for staying current.