What is in the data?
Load a CSV with Pandas, inspect units and missing values, select valid features and create train, validation and held-out test sets.
WORKSHOP 01 / THIRD-YEAR ENGINEERING
Machine Learning:
From Data to Decisions.
Make a prediction. Test the evidence. Explain the decision. Build a complete machine-learning experiment through one campus-energy case study.
Workshop reference: JIGYASA / WORKSHOP / 01
Basic Python is enough to begin. Prior ML practical experience is not assumed.
8 hours09:00–17:00, including breaks
405 minutesLearning and student work
300 minutesPractical blocks, including guidance
1 projectFrom CSV to a model card
THE WORK YOU WILL DO
Each experiment answers
a useful question.
How much energy?
Fit linear regression and compare its error with a training-mean baseline. Explain mean absolute error in kWh.
Will demand be high?
Train logistic regression, build a confusion matrix and interpret precision, recall and missed alerts.
Can we improve it?
Compare a decision tree in a guided experiment. Change depth and inspect training versus validation performance.
Can we trust the result?
Detect a deliberately leaked future measurement, correct the experiment, lock your approach and evaluate once on the held-out test.
Can someone else understand it?
Keep the notebook, comparison table, leakage finding and model card. Explain your own contribution in an individual exit response.
THE PROPOSED EIGHT-HOUR DAY
A clear pace.
Room to practise.
Short explanations lead into guided notebooks and paired work. Partners rotate the keyboard. The 300 practical minutes include demonstrations and troubleshooting; 75 minutes are reserved for breaks.
- 09:00–09:30
AI, ML and engineering judgement
Opening example and readiness check; distinguish rules, learned predictions and generative AI.
- 09:30–10:00
Understand the workflow
Features, targets, baselines and the roles of training, validation and test data.
- 10:00–10:45
Practical 1 / Explore and prepare
Load the dataset; inspect units and missing values; split before fitting preprocessing.
- 10:45–11:00
Morning break
15 minutes.
- 11:00–12:00
Practical 2 / Predict energy use
Linear regression, a mean baseline and interpreting prediction error.
- 12:00–12:45
Practical 3 / Classify demand
Logistic regression and high-demand alerts on validation data.
- 12:45–13:30
Lunch
45 minutes.
- 13:30–14:15
Practical 4 / Evaluate decisions
Confusion matrix, accuracy, precision, recall and validation-based threshold choices.
- 14:15–15:00
Practical 5 / Compare models
Decision-tree depth, overfitting and training versus validation performance.
- 15:00–15:15
Afternoon break
15 minutes.
- 15:15–16:15
Team challenge / Trust the result
Find leakage, correct the workflow, lock the chosen model and complete the final test and model card.
- 16:15–16:40
Evidence review
Selected team explanations; every pair submits its artifacts.
- 16:40–17:00
Individual assessment and next steps
Exit response, a seven-day extension and feedback.
BUILT AROUND YOUR CURRICULUM
Concepts from class.
Evidence from practice.
Designed for third-year engineering students who know basic Python. The sequence connects data preparation, supervised learning and evaluation to a complete practical workflow. We can adapt examples to the host institution’s syllabus before delivery.
AI & ML core
ML workflow, regression, classification and evaluation, with a runnable mini-project and evidence from held-out data.
ML elective connection
Preprocessing, validation, overfitting and decision-tree comparison. The host faculty can map these to local course outcomes.
What fits in one day?
A complete supervised-learning workflow is the focus. KNN, K-means, SVM, ensembles, deep learning and deployment are suitable follow-on topics; they are not additional full practicals in this day.
ARRIVE READY TO EXPERIMENT
Your workshop desk.
Before the day
Review Python variables, lists, functions and reading a table. A laptop or paired lab workstation with a CPU is sufficient. The delivery team will confirm the notebook environment in a technical preflight.
Tools and resources
Planned tools: Python, NumPy, Pandas, Matplotlib and scikit-learn, using Jupyter or Colab where access permits. Six guided notebooks, the synthetic dataset, learner notes, slides and revision podcasts are available in the learning workspace.
Resource status: the reusable teaching pack is available. Each host confirms its event date, lab setup and delivery arrangements.
What you take home
A runnable notebook, baseline/model comparison, confusion matrix, leakage finding and a short model card. A seven-day extension asks you to change one valid feature or model setting and justify the result.
For the college
Proposed capacity: up to 60 students in pairs, supported by one lead facilitator and two mentors. Lab capacity, support staff, accessibility needs and date are confirmed with the institution.
PROGRAMME FACULTY / AKASH CHATAKE
Bring the workshop
to your cohort.
Proposed for third-year engineering students. Dates and institutional delivery are subject to agreement. This page does not open enrolment or issue completion certificates.