AI & Machine Learning

Builds fluent, production-grade Python and array/dataframe skills so students can manipulate real-world datasets from day one.

• Python syntax, data types, control flow & functions for data work
• OOP fundamentals: classes, objects, inheritance for reusable pipelines
• File I/O and working with CSV/JSON/Excel data sources
• List/dict comprehensions and lambda, map, filter, reduce
• NumPy ndarrays: creation, indexing, broadcasting & vectorized ops
• NumPy linear algebra essentials: dot products, matrix ops, reshaping
• Pandas Series & DataFrame: indexing, slicing, filtering
• GroupBy, aggregation, pivot tables & merge/join operations
• Reading/writing data at scale and handling large files with chunking
• Jupyter Notebook workflow, magic commands & reproducible analysis
• Data visualization basics with Matplotlib: line, bar, scatter, histogram
• Writing clean, PEP8-compliant, well-documented Python code
• Virtual environments & dependency management with pip/conda
• Debugging and profiling Python code for performance

Turns messy, real-world data into model-ready datasets through systematic cleaning, exploration and statistical reasoning.

• Exploratory Data Analysis workflow: understand, question, visualize
• Handling missing values: deletion, mean/median/mode & KNN imputation
• Outlier detection using IQR, Z-score & visualization techniques
• Univariate, bivariate & multivariate analysis with Seaborn
• Data cleaning: deduplication, type coercion & inconsistent labels
• Encoding categorical variables: one-hot, label & ordinal encoding
• Feature scaling: standardization, min-max & robust scaling
• Handling imbalanced datasets: SMOTE, undersampling & class weights
• Correlation analysis, heatmaps & multicollinearity checks
• Working with date/time features and time-based aggregation
• Statistical foundations: distributions, hypothesis testing & p-values
• Data storytelling: crafting narrative dashboards from EDA findings
• Detecting data leakage and train/test contamination early
• Automated EDA reporting with pandas-profiling/Sweetviz-style tools

Equips students to build, tune and reason about classical supervised and unsupervised ML models used in real interviews and projects.

• Supervised vs unsupervised vs reinforcement learning paradigms
• Linear & polynomial regression with gradient descent intuition
• Logistic regression for binary and multiclass classification
• K-Nearest Neighbors & distance-based learning
• Decision trees: entropy, Gini impurity & pruning strategies
• Ensemble methods: Random Forest, Bagging & Boosting (XGBoost/LightGBM)
• Support Vector Machines with kernel tricks
• Naive Bayes classifiers for text and categorical data
• K-Means, hierarchical & DBSCAN clustering for unsupervised learning
• Dimensionality reduction with PCA & t-SNE
• Recommendation systems: collaborative & content-based filtering
• Bias-variance tradeoff and overfitting/underfitting diagnosis
• Building end-to-end ML pipelines with Scikit-learn Pipeline API
• Model persistence and versioning with joblib/pickle

Sharpens model quality and interview-readiness by teaching how to engineer signal-rich features and rigorously validate performance.

• Feature creation: interaction terms, polynomial & domain-driven features
• Automated feature selection: filter, wrapper & embedded methods (RFE, Lasso)
• Handling cyclical features: encoding hour/day/month as sine-cosine
• Binning, discretization & target encoding for high-cardinality categoricals
• Cross-validation strategies: K-Fold, Stratified K-Fold & time-series splits
• Hyperparameter tuning with GridSearchCV, RandomizedSearchCV & Optuna
• Classification metrics: accuracy, precision, recall, F1 & ROC-AUC
• Regression metrics: MAE, MSE, RMSE & R-squared interpretation
• Confusion matrix analysis & threshold tuning for business goals
• Detecting and correcting overfitting via regularization (L1/L2)
• Learning curves and diagnosing high-bias vs high-variance models
• Feature importance interpretation with SHAP & permutation importance
• A/B testing fundamentals for validating model impact in production
• Building a reusable model evaluation & reporting framework

Prepares students to design and train neural networks for vision and structured-data tasks using industry-standard frameworks.

• Perceptrons, activation functions & the biological neuron analogy
• Forward propagation, backpropagation & the chain rule intuition
• Loss functions and optimizers: SGD, Momentum, RMSProp & Adam
• Building feedforward neural networks with Keras/TensorFlow
• Building the same networks with PyTorch tensors & autograd
• Regularization techniques: dropout, batch normalization & early stopping
• Convolutional Neural Networks: filters, pooling & feature maps
• Transfer learning with pretrained CNNs (ResNet, VGG, EfficientNet)
• Image classification & object detection basics (YOLO overview)
• Recurrent Neural Networks, LSTMs & GRUs for sequence data
• Hyperparameter tuning for deep nets: learning rate schedules & callbacks
• GPU acceleration and mixed-precision training fundamentals
• Model checkpointing, TensorBoard/experiment tracking
• Deploying trained models for inference with TensorFlow Serving/TorchScript

Gives students hands-on command of modern NLP and generative AI so they can build and reason about LLM-powered applications.

• Text preprocessing: tokenization, stemming, lemmatization & stopwords
• Bag-of-Words, TF-IDF & classical text classification pipelines
• Word embeddings: Word2Vec, GloVe & contextual embeddings
• Sequence models for NLP: RNN/LSTM limitations vs Transformers
• Self-attention & the Transformer architecture end-to-end
• BERT and encoder models for classification, NER & QA tasks
• GPT-style decoder models & autoregressive text generation
• Fine-tuning pretrained models with Hugging Face Transformers/Trainer
• Prompt engineering: zero-shot, few-shot & chain-of-thought prompting
• Retrieval-Augmented Generation: embeddings, vector stores & chunking
• Parameter-efficient fine-tuning: LoRA/QLoRA & instruction tuning basics
• Building LLM apps with the OpenAI API & function/tool calling
• Evaluating generative outputs: hallucination checks & RAG faithfulness
• Responsible AI: bias, toxicity filtering & prompt-injection awareness

Turns trained models into shippable products by teaching version control, packaging, APIs and deployment workflows employers expect.

• Git fundamentals: init, commit, branch, merge & conflict resolution
• Collaborative workflows: pull requests, code review & GitHub Actions basics
• Structuring an ML project repo: src, tests, configs & requirements
• Serializing models with pickle/joblib and versioning artifacts
• Building REST APIs for model inference with Flask
• Request validation, error handling & API documentation for ML services
• Containerizing ML applications with Docker and writing Dockerfiles
• Environment reproducibility with requirements.txt/Docker layers
• Basic CI/CD concepts for automated testing and deployment
• Model monitoring: drift detection, logging & performance alerts
• Deploying to cloud platforms (Render/Heroku/AWS/GCP basics)
• Building a simple front-end demo (Streamlit/HTML) for model showcase
• Writing unit tests for data pipelines and model inference code
• Capstone project: end-to-end ML system from data to deployed API