Skip to content

CS 105: Practical Machine Learning
Course Code CS 105
Course Name Practical Machine Learning
Department Computer Science
Semester Offered Even (Term 2 - India)
Tuition Hours 30 hours
Course Level Intermediate
Pre-requisite CS 103: Data Structures & Algorithm Design
Co-requisite CS 104: Data Engineering for Emerging Markets
Course Objective Machine learning is often taught as theory first. That is a mistake if your goal is to build real systems.

This course flips that approach. Students start with real problems and real data, and learn models as tools to solve them. The focus is not on deriving equations, but on understanding when to use which model, how to train it, and how to make it work in production.

Students will implement core machine learning techniques such as regression, classification, and tree-based methods using practical tools like scikit-learn. They will also learn to use pre-trained models and transfer learning to move faster.

By the end of the course, students will be able to build, evaluate, and deploy machine learning systems that solve real business problems, directly contributing to their Term 2 enterprise AI project.
Course Philosophy This course emphasizes
  • Use before theory
  • Models as tools, not as ends
  • Iteration over perfection
Students will learn machine learning by building working systems quickly, then improving them. The goal is not to know every algorithm, but to know how to get something working and improve it based on feedback and data.
Course Learning Outcomes Upon successful completion of this course, students will be able to:
  • Understand and apply core ML techniques such as regression and classification.
  • Train and evaluate models using real-world datasets.
  • Use scikit-learn effectively to build end-to-end ML pipelines.
  • Select appropriate models based on problem type and data characteristics.
  • Improve model performance through feature engineering and tuning.
  • Use tree-based models like decision trees and random forests for practical problems.
  • Leverage pre-trained models and transfer learning to accelerate development.
  • Deploy ML models into simple production environments for real-world use.
Course Author Sagar Udasi
MSc Statistics and Data Science with Computational Finance from The University of Edinburgh.
Contact: sagar.l.udasi@gmail.com
Course Organiser TBD
Details will be updated before course commencement.
No. Lecture Title Concepts Covered Lecture Objective
01 What Problem Are You Actually Solving? ML problem framing, supervised learning Teach students to map business problems to ML tasks.
02 From Raw Data To Training Data Feature selection, preprocessing Prepare datasets for model training.
03 Your First Model That Actually Predicts Linear regression, training basics Build the first working predictive model.
04 When Predictions Go Wrong Bias, variance, errors Understand why models fail and how to improve them.
05 Classification: Decisions, Not Numbers Logistic regression, classification basics Handle categorical prediction problems.
06 Measuring If Your Model Is Useless Evaluation metrics, accuracy, precision, recall Teach proper evaluation of models.
07 Overfitting: When Your Model Lies To You Overfitting, generalization Build intuition about model reliability.
08 Trees That Think Like Humans Decision trees Introduce interpretable models.
09 When One Tree Isn’t Enough Random forests, ensemble methods Improve performance using multiple models.
10 Feature Engineering Is Where The Magic Happens Feature creation, transformation Improve model performance using better data.
11 Pipelines: Making It Repeatable ML pipelines, workflows Structure ML processes for consistency.
12 Using Models You Didn’t Train Pre-trained models, transfer learning Speed up development using existing models.
13 Working With Real Business Data Case study datasets Apply ML to realistic problems.
14 When Data Is Not Clean Handling noise, missing values Build robustness in ML systems.
15 Speed vs Accuracy: What Matters? Trade-offs in model design Align model performance with business goals.
16 From Notebook To API Serving models, basic deployment Turn models into usable services.
17 Integrating ML Into Products ML + backend systems Connect ML outputs to real applications.
18 Monitoring Models In The Wild Drift, feedback loops Ensure models remain useful over time.
19 Case Study: ML For An Enterprise Problem End-to-end ML system Directly connect to Term 2 capstone.
20 Demo Day: Does Your Model Create Value? Presentations, evaluation Students present deployed ML solutions.
Component Weightage
Weekly ML Assignments (5 total) 30%
Model Building Project 20%
Final Project: Deployed ML Solution 30%
Viva + Model Evaluation Review 20%
Type Resource Provider
Lecture Machine Learning Crash Course Google
Lecture Applied Machine Learning Columbia University
Reading Hands-On Machine Learning with Scikit-Learn & TensorFlow Aurélien Géron
Reading Pattern Recognition and Machine Learning Christopher Bishop
Practice Kaggle kaggle.com
Documentation Scikit-learn Documentation scikit-learn.org