AI and Machine Learning Engineering

Topics

• What AI is and what it is not — separating capability from marketing
• Types of AI — ANI, AGI, ASI — what exists today, what does not, why the distinction matters
• History and evolution — Turing, AI winters, deep learning, transformers, the generative AI moment
• Real-world AI applications — healthcare, finance, EdTech, logistics, Tamil Nadu manufacturing context
• AI vs Machine Learning vs Deep Learning — the actual relationships
• The 2026 AI landscape and what it means for careers
• Careers in AI — ML engineer, data scientist, AI product manager, MLOps engineer, AI application developer

Hands-on

• Structured exploration of ChatGPT, Claude, and Google AI tools — probing capabilities, failure modes, and differences

Topics

• Python fundamentals review — variables, data types, loops, functions, scope
• Data structures for AI — lists, tuples, dictionaries, sets
• File handling — reading CSV, JSON, and text files, writing outputs
• OOP basics — classes, instances, inheritance to understand ML library APIs
• NumPy — arrays, broadcasting, indexing, reshaping
• Pandas — Series and DataFrame, filtering, groupby, missing values
• Writing clean, readable AI code — naming, modular functions, notebooks vs scripts

Hands-on

• Data manipulation on real datasets (sales, student performance, weather)
• Data pipeline from raw CSV to clean analysis-ready DataFrame

AI integration

• Using Claude and ChatGPT to explain NumPy and Pandas errors and suggest operations, with verification

Topics

• Linear algebra for ML — vectors, matrices, matrix multiplication, dot products
• Probability fundamentals — conditional probability, Bayes' theorem
• Statistics — mean, variance, correlation, distributions (normal, binomial, uniform)
• The gradient concept — derivatives, loss minimisation, gradient descent
• Overfitting and underfitting — bias-variance tradeoff

Hands-on

• Implement linear algebra and probability operations in NumPy
• Hand-coded gradient descent on a simple function

Topics

• Data cleaning — missing values, duplicates, inconsistent formatting, outliers
• Exploratory Data Analysis — distributions, correlations, feature relationships
• Feature engineering basics — encoding, scaling, normalisation
• Matplotlib and Seaborn — core plots, heatmaps, pairplots, violin plots
• Communicating findings to non-technical stakeholders

Hands-on

• Full EDA on sales, student performance, and healthcare datasets
• Visualisation and written findings in a reusable notebook format

Topics

• ML workflow, train/test splits, cross-validation, bias-variance tradeoff
• Regression — linear, polynomial, logistic, regularisation (Ridge, Lasso)
• Classification — KNN, decision trees, random forests, SVMs, Naive Bayes
• Unsupervised learning — K-Means, hierarchical clustering, PCA
• Evaluation — accuracy, precision, recall, F1, ROC-AUC, regression metrics

Hands-on

• Classification and regression projects (grades, loan default, housing, churn)
• Interpretation documents for each model (what it learned, where it fails, next steps)

Topics

• Neural network fundamentals — perceptron, layers, activations, forward and backward propagation
• Loss functions and optimisers — cross-entropy, MSE, SGD, Adam, learning rate
• CNNs for images — convolutions, pooling, architecture patterns, transfer learning
• TensorFlow and Keras — model definition, training, callbacks, regularisation

Hands-on

• Build and train an image classifier from scratch
• Implement transfer learning on a different image dataset using a pre-trained model

Topics

• NLP pipeline — preprocessing, tokenisation, stemming/lemmatisation
• Bag of Words, TF-IDF, classical text classification and sentiment analysis
• Word embeddings — Word2Vec, GloVe; why they matter
• Transformers and attention (conceptual), pre-trained models (BERT, GPT)
• Hugging Face basics — pipelines, fine-tuning on small datasets

Hands-on

• Sentiment analysis for product reviews
• Text classification for a real categorisation problem
• Using a Hugging Face model for classification through code

Topics

• Saving and loading models — Pickle, Joblib, ONNX (awareness)
• FastAPI model APIs — input validation, error handling, deployment readiness
• Streamlit apps — quick interactive ML demos and dashboards
• Deployment platforms — Render, Railway, Hugging Face Spaces
• Model monitoring basics — data and concept drift awareness
• MLOps introduction — MLflow and DVC at awareness level

Hands-on

• Deploy a classification model as a FastAPI endpoint
• Build a Streamlit dashboard around a regression model
• Publish a complete ML project to GitHub with README and live demo

Format

• End-to-end AI/ML project: problem definition, data, EDA, modelling, evaluation, deployment, documentation
• Project scoped with your trainer during Module 9 to match your interests and target industry

Example projects

• Education: student performance prediction system with risk scores and teacher dashboard, including fairness reflection.
• Healthcare: disease risk screening tool with responsible AI framing and a documented model card.
• Business: customer churn prediction with business value quantification and an explorable Streamlit interface.
• Media: news classifier and summariser using ML plus LLM summarisation with RAG grounding.
• Agriculture: crop yield prediction for Tamil Nadu districts, including uncertainty and data quality analysis.

Assessment

20-minute viva — you present your project, walk through your methodology, explain model choice, and answer questions on what you would do differently and what breaks if input data shifts.