Covers measurement error, uncertainty, bias, validity, reliability, missing values, and data quality checks. Learners build the habit of asking where data came from and what it can safely support.

Shows how spreadsheets, CSV files, delimiters, encodings, and tidy table structure work. Learners clean a small spreadsheet and turn it into an analysis-ready table.

Builds the mental model behind variables, conditions, loops, functions, and reproducible steps. Learners write simple procedures for counting, filtering, and transforming data.

Covers Python syntax, data types, functions, errors, files, and simple scripts for data tasks. Learners write small programs that load data, process rows, and produce results.

Uses NumPy, pandas, and common plotting libraries for real data analysis. Learners filter, group, join, summarize, and visualize datasets in Python.

Covers Jupyter notebooks, virtual environments, package installation, file paths, and basic command-line use. Learners set up a working data science project folder that can run again later.

Shows how Git tracks changes, supports collaboration, and protects project history. Learners commit notebooks, scripts, data notes, and reports in a clean repository.

Traces data science from statistics, databases, business intelligence, scientific computing, and machine learning into today’s cloud and AI toolchains. The chapter explains why the field combines evidence, computation, products, and decisions.

Covers how to turn a vague request into a clear analytical question, success measure, scope, and deliverable. Learners write project briefs that separate what can be answered from what needs more data.

Shows how metrics connect data work to real decisions, incentives, tradeoffs, and domain knowledge. Learners design useful metrics and spot cases where a metric can mislead people.

Covers consent, harm, representation, surveillance, sensitive attributes, and respectful data use. Learners review a data project for ethical risks before any modeling begins.

Covers center, spread, ranks, percentiles, correlation, and summary tables. Learners describe a dataset accurately without jumping to unsupported conclusions.

Builds probability with events, conditional probability, independence, Bayes’ rule, and expected value. Learners use probability to reason about uncertainty in everyday and data science settings.

Covers populations, samples, sampling bias, random assignment, surveys, and observational studies. Learners judge whether a dataset can represent the group it claims to describe.

Covers common distributions, random variables, variance, standard error, and simulation. Learners use distributions to model real variation and check whether results are surprising.

Covers estimators, confidence intervals, standard errors, bootstrap methods, and the logic of inference. Learners quantify uncertainty instead of reporting single numbers as if they were exact.

Covers null hypotheses, p-values, statistical power, multiple comparisons, and practical significance. Learners run tests responsibly and explain what the results do and do not prove.

Covers priors, likelihoods, posteriors, credible intervals, and Bayesian updating. Learners solve small uncertainty problems where prior knowledge and new evidence must be combined.

Covers vectors, matrices, dot products, projections, matrix factorization, and high-dimensional data. Learners connect these ideas to regression, embeddings, recommender systems, and deep learning.

Covers rates of change, gradients, loss functions, and numerical optimization. Learners see how models improve by minimizing error rather than by magic.

Covers control groups, randomization, blocking, confounding, power planning, and measurement choices. Learners design experiments that can support stronger claims about cause and effect.

Covers online experiments, treatment assignment, guardrail metrics, sample size, peeking risks, and rollout decisions. Learners plan and analyze an A/B test for a product or service change.

Covers causal graphs, confounders, interventions, matching, difference-in-differences, instrumental variables, and causal assumptions. Learners separate prediction questions from cause-and-effect questions.

Covers data collection from files, databases, forms, sensors, public datasets, and APIs. Learners request, document, and store data while keeping provenance clear.

Covers HTML structure, scraping tools, rate limits, robots.txt, terms of service, and respectful collection practices. Learners collect simple web data without treating the public internet as a free-for-all.

Covers SELECT queries, filtering, sorting, grouping, joins, subqueries, window functions, and query debugging. Learners answer real analytical questions directly from relational data.

Covers primary keys, foreign keys, normalization, fact tables, dimension tables, and star schemas. Learners design database structures that make analysis reliable and efficient.

Covers document stores, key-value stores, graph databases, and when nonrelational storage fits data science work. Learners choose storage patterns based on access needs rather than trends.

Covers type conversion, duplicate records, inconsistent categories, units, dates, text cleanup, and reproducible cleaning scripts. Learners turn messy raw data into a trustworthy working dataset.

Covers missingness patterns, deletion, simple imputation, model-based imputation, and the risks of hiding missing data. Learners decide how to handle missing values without inventing false certainty.

Covers outlier detection, heavy-tailed data, robust summaries, winsorization, and sensitivity checks. Learners decide when an unusual value is an error, a rare event, or the most important part of the story.

Covers the practical cycle of summarizing, visualizing, slicing, questioning, and checking assumptions. Learners produce an exploratory notebook that records both findings and doubts.

Covers visual encodings, chart choice, scales, color, uncertainty, and accessibility. Learners create charts that reveal patterns without distorting the data.

Covers dashboards, filters, drilldowns, refresh schedules, stakeholder needs, and common business intelligence patterns. Learners build a dashboard that supports repeated decisions rather than one-time curiosity.

Covers narrative structure, evidence, caveats, executive summaries, and audience-specific explanations. Learners turn analysis into a report or presentation that helps people act responsibly.

Follows a complete analysis from question framing through data acquisition, cleaning, exploration, statistics, visualization, and final recommendation. Learners practice the full rhythm of a professional data analysis project.

Covers the idea of learning from labeled examples, training data, test data, features, targets, loss, and generalization. Learners train their first predictive models and compare them to simple baselines.

Covers linear regression, regularization, diagnostics, residuals, interactions, and prediction intervals. Learners build regression models for numerical outcomes and explain their limits.

Covers logistic regression, decision trees, random forests, gradient boosting, class imbalance, and probability calibration. Learners build models that predict categories while respecting real-world error costs.

Covers train-test splits, cross-validation, leakage, confusion matrices, ROC curves, precision, recall, and model selection. Learners judge models by the decision they must support, not just by a single score.

Covers encoding categories, scaling, transformations, date features, text features, aggregations, and feature leakage. Learners create features that improve models while staying faithful to what would be known at prediction time.

Covers permutation importance, partial dependence, SHAP-style explanations, surrogate models, and explanation limits. Learners communicate why a model made predictions and where those explanations may fail.

Covers pattern discovery without labels, distance measures, similarity, representation, and evaluation challenges. Learners use unsupervised methods carefully when there is no clear target variable.

Covers k-means, hierarchical clustering, density-based clustering, cluster validation, and cluster interpretation. Learners group customers, documents, or observations while avoiding false stories about arbitrary clusters.

Covers principal component analysis, manifold methods, embeddings, visualization of high-dimensional data, and information loss. Learners compress complex data while keeping the purpose of the compression clear.

Covers statistical rules, isolation forests, reconstruction error, rare-event evaluation, and alert fatigue. Learners build anomaly detection systems for fraud, quality control, operations, or security use cases.

Covers timestamps, seasonality, trend, autocorrelation, lag features, backtesting, and forecasting horizons. Learners forecast demand, traffic, revenue, or sensor readings with honest uncertainty.

Covers collaborative filtering, matrix factorization, ranking metrics, cold-start problems, and feedback loops. Learners build recommendation models and evaluate whether they improve user experience or only chase clicks.

Covers nodes, edges, centrality, communities, graph features, and graph-based prediction. Learners analyze networks such as transactions, social links, citations, or supply chains.

Covers coordinates, projections, spatial joins, spatial aggregation, maps, and geographic bias. Learners analyze location data without making misleading maps or distance calculations.

Covers tokenization, text cleaning, bag-of-words, TF-IDF, topic models, sentiment analysis, and text classification. Learners build useful text models before moving into modern neural language systems.

Covers neural networks, backpropagation, activation functions, regularization, optimizers, and training curves. Learners train small neural models and diagnose underfitting, overfitting, and unstable training.

Covers image data, convolutional networks, image classification, object detection ideas, augmentation, and evaluation. Learners build computer vision models while noticing dataset bias and labeling quality.

Covers agents, environments, rewards, policies, value functions, exploration, and offline evaluation risks. Learners see when reinforcement learning fits decision problems and when simpler methods are safer.

Covers GPU hardware concepts, batching, memory limits, acceleration libraries, and cost-aware training. Learners decide when GPUs are useful and how to avoid wasting compute.

Covers attention, transformer architecture, embeddings, pretraining, transfer learning, and transformer-based prediction tasks. Learners connect modern language, vision, and multimodal systems to their statistical roots.

Covers foundation models, large language models, tokens, context windows, instruction tuning, hallucination, and evaluation. Learners use LLMs as powerful but fallible tools for data and text work.

Covers prompt structure, tool use, code generation, notebook assistance, synthetic examples, and verification habits. Learners use AI assistants to speed up work without outsourcing judgment.

Covers document chunking, embeddings, vector search, retrieval quality, grounding, citations, and failure modes. Learners build a retrieval-augmented generation system that answers from a controlled knowledge base.

Covers supervised fine-tuning, parameter-efficient methods, instruction data, evaluation sets, and cost tradeoffs. Learners decide when adapting a model is better than prompting or retrieval alone.

Covers diffusion model ideas, image generation, tabular synthetic data, data augmentation, and privacy risks. Learners judge when generated data helps and when it creates new bias or leakage.

Covers modular code, functions, packages, logging, configuration, code review, and maintainable project structure. Learners move from one-off notebooks toward reusable data science software.

Covers unit tests, data tests, schema checks, expectation suites, and regression tests for analysis results. Learners catch broken assumptions before they reach dashboards, models, or decision makers.

Covers containers, images, Dockerfiles, dependency pinning, reproducible builds, and portable execution. Learners package a data project so it can run on another machine or in the cloud.

Covers pods, jobs, services, resource requests, scaling, and when Kubernetes is worth the complexity. Learners see how data and machine learning workloads run in production clusters.

Covers scheduled jobs, task dependencies, retries, backfills, orchestration tools, and operational handoffs. Learners build pipelines that run reliably after the original notebook is closed.

Covers partitioning, parallel processing, Spark-style dataframes, shuffle costs, and cluster execution. Learners process datasets that are too large or too slow for a single machine workflow.

Covers managed analytical databases, columnar storage, query optimization, workload management, and cost control. Learners use cloud warehouses for scalable SQL analytics without treating them like ordinary spreadsheets.

Covers object storage, raw zones, curated zones, open file formats, and governance problems in large shared data stores. Learners organize data lakes so they remain useful instead of becoming dumping grounds.

Covers lakehouse tables, ACID transactions on object storage, schema evolution, time travel, and mixed SQL and machine learning workloads. Learners see why lakehouse architecture became a major pattern for modern data platforms.

Covers events, message queues, stream processing, windows, late data, and real-time metrics. Learners design analytics that respond to data as it arrives instead of waiting for batch reports.

Covers transformation projects, tested SQL models, metric definitions, semantic layers, and collaboration between analysts and engineers. Learners build trusted business datasets that many teams can reuse.

Covers metadata, lineage, ownership, data catalogs, discovery, and impact analysis. Learners trace where data came from, who owns it, and what will break if it changes.

Covers batch scoring, real-time APIs, model packaging, latency, throughput, fallback behavior, and rollback plans. Learners turn trained models into services or jobs that other systems can use.

Covers experiment tracking, model registries, reproducible training runs, approval stages, and release management. Learners manage machine learning work as a controlled lifecycle rather than a pile of files.

Covers data drift, concept drift, performance tracking, alerting, logging, observability, and incident response. Learners keep models useful after deployment and know when to retrain, pause, or retire them.

Covers access control, secrets, encryption, secure sharing, dependency risk, and attacks on data workflows. Learners protect analytical systems from leaks, tampering, and unsafe shortcuts.

Covers personal data, de-identification, retention, consent records, cross-border transfer, GDPR-style rights, HIPAA-style duties, and sector rules. Learners plan data work that respects legal and organizational obligations.

Covers aggregation, masking, differential privacy, federated learning, secure enclaves, and privacy-utility tradeoffs. Learners choose privacy-preserving methods when ordinary access controls are not enough.

Covers measurement bias, representation gaps, subgroup performance, fairness metrics, proxy variables, and mitigation tradeoffs. Learners test whether data products work differently for different groups.

Covers model cards, data sheets, approval processes, audit trails, human oversight, risk tiers, and AI policy expectations. Learners document and govern data science systems before they affect people at scale.

Follows a data product from request intake through data contracts, pipeline design, modeling, validation, deployment, monitoring, user feedback, and iteration. Learners practice the full workflow used by teams that operate data products over time.

Guides learners through a complete portfolio project with a clear question, documented data, tested code, analysis, model or dashboard, written findings, and deployment or reproducible handoff. The final artifact shows practical skill rather than only course completion.

Covers roles such as analyst, data scientist, machine learning engineer, analytics engineer, data engineer, and applied AI practitioner. Learners plan entry paths, portfolio proof, interview preparation, certifications when useful, communities, journals, conferences, and habits for staying current.