Guided Labs

Guided Labs#

“Because Theory Without Practice Is Like a Pizza Without Cheese.”#

🧭 Welcome to the Practice Arena#

You’ve been reading, coding, laughing, and maybe even crying your way through this course. Now it’s time to get your hands dirty — not with soil, but with CSV files, feature engineering, and random seeds. 🌱

These guided labs are your training dojo — small, focused challenges that help you connect:

🤖 Machine learning theory → 💼 Business application → 💬 “Hey boss, look what I automated!”

Each lab is structured like a mini adventure:

  1. Setup – grab your notebook and data.

  2. Task – implement key ML logic.

  3. Analysis – interpret results like a consultant, not a robot.

  4. Reflection – what would you tell a CEO about it?

🧰 Tools of the Trade#

You’ll need:

  • Python (obviously)

  • pandas, numpy, matplotlib, scikit-learn

  • PyTorch (for the deep learning ones)

  • And your best debugging face 😤

Each lab can be:

  • 🧠 Run online via JupyterLite (no installation headaches)

  • 🧑‍💻 Launched in Google Colab

  • 📥 Downloaded as a .ipynb notebook to flex your offline powers

🧩 Lab Topics Overview#

Lab

Description

Goal

1. Regression: Sales Forecasting

Predict sales trends using linear regression

Connect model outputs to revenue KPIs

2. Classification: Customer Churn

Predict which customers will leave

Learn business impact of false positives

3. Time Series Forecasting

Build demand forecasts for inventory

Manage supply chain planning

4. NLP Sentiment Analysis

Extract opinions from customer feedback

Translate text data into strategy

5. Optimization Lab

Compare Gradient Descent variants

Learn training stability & cost tradeoffs

6. Deep Learning Lab

OCR PDFs using CNNs

Apply neural networks to unstructured data

7. LLM Lab

Fine-tune a small transformer

See how AI assistants can automate insight generation

🕹️ Format Example – “Regression Lab: Sales Forecasting”#

Goal: Predict next quarter’s sales using linear regression and discuss how business KPIs might be influenced.

Steps:

  1. Load the dataset (sales_data.csv)

  2. Split into train/test

  3. Fit a regression model

  4. Plot actual vs predicted sales

  5. Calculate metrics (MAE, RMSE, R²)

  6. Write a short paragraph for your “boss” explaining what it means

🧠 Hint: Don’t just say “RMSE = 12.3”. Say “Our forecasts are off by about 12 units — meaning we overstocked by ~5% last quarter.”

💼 Boss Mode Challenge (Optional but Fun)#

Each lab ends with a Boss Mode Challenge — an open-ended extension to show off.

Example:

“Your model works great on average. Now test it by region and identify where it fails. Bonus: create a Streamlit dashboard showing results by location!”

If you complete a few of these, you’ll have something powerful for your portfolio. 🚀

🧮 Evaluation Criteria#

Criteria

Description

Accuracy

Does your model actually work?

Business Insight

Can you connect numbers to decisions?

Communication

Clear, concise explanations

Creativity

Unique methods or visualizations

Reproducibility

Notebook runs without breaking (we hope)

🧘 The Spirit of the Labs#

This isn’t about “perfect answers.” It’s about thinking like a data scientist in a business context.

You’ll fail, fix, learn, and laugh. Sometimes all in the same cell.

💬 “Data science is 90% cleaning, 10% debugging, and 100% pretending you planned it that way.”

# Your code here
Contents