Data Science & AI
From messy tables to a model you can defend — plus a working chatbot.
Syllabus
Models are the last ten percent. This six-month course starts where real work starts: dirty tables, missing values, and questions that are not yet queries. You learn Python for analysis, SQL for the warehouse-shaped problems, and Pandas until grouping and joining are reflexes. Visualisation is a sentence, not a rainbow. Then supervised learning: train, test, leak, and the humility of a baseline. You will predict internal marks from study features you actually collected, classify images with a pipeline you can re-run, and build a retrieval chatbot over a small knowledge set so “AI” is a system, not a screenshot. Mentors challenge metrics that look good and mean nothing. Classroom and live-online batches share notebooks, reviews, and a final viva where you explain a wrong answer. You leave able to tell a faculty member what the model does, what it misses, and which three projects prove it.
What you leave with
- Clean and query data before you touch a model
- Report metrics a sceptic would accept
- Explain a miss as clearly as a hit
Three projects
01
Student performance model
Predict internals from features you collected. Beat a baseline, then explain the misses.
02
Image classifier pipeline
A re-runnable training notebook with a held-out test set and a short error report.
03
AI chatbot assistant
Retrieval over a small knowledge set — answers you can trace, not a black box demo.
Related showcase: AI Chatbot Assistant