10 - Machine Learning, Automation, Version Control, and Responsible AI — Topics & Learning Outcomes

Module Topics

Foundations of Supervised Machine Learning

• What Is Supervised Learning? — Supervised learning is a machine learning paradigm in which a model is trained on labeled input-output pairs to learn a mapping from features to targets.
• Regression vs. Classification — Regression predicts a continuous numerical output, while classification predicts a discrete category or class label.
• Features, Labels, and Training Data — Features are the measurable input variables used to make predictions, while labels are the corresponding known outputs used during training.
• Bias-Variance Trade-off — The bias-variance trade-off describes the tension between a model that is too simple (high bias) and one that is too complex (high variance).
• Model Evaluation Metrics — Selecting appropriate evaluation metrics is essential for understanding how well a model performs on the intended engineering task.

Machine Learning with scikit-learn

• The scikit-learn API Design — scikit-learn provides a consistent estimator API in which every model object exposes fit, predict, and score methods, making it straightforward to swap algorithms.
• Data Preprocessing and Pipelines — Raw engineering data often requires scaling, encoding, or imputation before it can be fed into a machine learning model, and scikit-learn pipelines chain these steps with the estimator.
• Regression Models in scikit-learn — scikit-learn offers a range of regression estimators suitable for engineering data, from linear models to ensemble methods.
• Classification Models in scikit-learn — Classification estimators in scikit-learn assign input samples to discrete classes and are evaluated using metrics such as accuracy, precision, and recall.
• Hyperparameter Tuning — Hyperparameters are model configuration settings that are not learned from data and must be set before training; tuning them improves model performance.
• Model Interpretability and Feature Importance — Understanding which features drive model predictions is critical in engineering applications where decisions have physical and safety consequences.

Engineering Automation and Parameter Sweep Pipelines

• Why Automate Engineering Workflows? — Automation replaces manual, error-prone repetition with reliable, reproducible scripts that can execute complex sequences of engineering tasks consistently.
• Designing a Parameter Sweep — A parameter sweep systematically varies one or more input parameters across a defined range and records the corresponding outputs to map system behavior.
• Integrating ML Models into Automation Pipelines — Trained machine learning models can be embedded in automation pipelines to provide fast surrogate predictions in place of expensive simulations or experiments.
• Handling Large Result Sets — Parameter sweeps can generate large volumes of output data that require efficient storage, querying, and summarization strategies.
• Scheduling and Running Long Jobs — Engineering automation scripts that take significant time to run benefit from progress tracking, logging, and the ability to resume after interruption.

Git and GitHub Version Control

• Why Version Control Matters for Engineers — Version control systems track every change made to a codebase, enabling engineers to review history, revert mistakes, and collaborate without overwriting each other's work.
• Core Git Concepts and Commands — Git operates through a staging area and a local repository; understanding the add, commit, status, and log commands forms the foundation of daily Git use.
• Branching and Merging — Branches allow engineers to develop features or experiments in isolation from the main codebase, and merging integrates completed work back into the main branch.
• Working with GitHub Remote Repositories — GitHub hosts remote copies of Git repositories, enabling cloud backup, sharing, and team collaboration through push and pull operations.
• Pull Requests and Code Review — Pull requests on GitHub are a formal mechanism for proposing changes, inviting peer review, and merging approved work into a shared branch.
• Repository Hygiene and .gitignore — Good repository hygiene ensures that only relevant source files are tracked, avoiding clutter from generated outputs, credentials, or environment-specific files.

Responsible AI Integration in Engineering Practice

• The Role of AI Tools in Engineering Workflows — AI coding assistants and generative tools can accelerate engineering software development, but they introduce new responsibilities around verification and professional judgment.
• Verification of AI-Generated Code — Every piece of code produced by an AI assistant must be critically reviewed, tested, and validated before it is used in any engineering computation or product.
• Recognizing Hallucinations and Errors — AI language models can confidently produce incorrect code, fabricated function names, or physically unreasonable results—a phenomenon known as hallucination.
• Professional Engineering Accountability — Licensed and professional engineers bear legal and ethical responsibility for the work they produce and cannot transfer that responsibility to an AI tool.
• Bias, Fairness, and Equity in Engineering ML Systems — Machine learning models trained on historical data can inherit and amplify societal biases, leading to unfair or discriminatory outcomes in engineering systems that affect people.
• Transparent Documentation and Reproducibility — Responsible AI practice requires that models, datasets, training procedures, and validation results be documented thoroughly so that others can understand, audit, and reproduce the work.
• Integrating Responsible AI Habits into Daily Practice — Responsible AI is not a separate checklist but a set of habits—verification, documentation, skepticism, and ethical reflection—that should be embedded in every engineering workflow.

Student Learning Outcomes

By the end of this module, students will be able to:

Course Outcomes (reference)