AI Specialization Program

• Soft AI Academy
AI Specialization Program
•
Technical Tracks
AI Engineering & Data
•
Week 2

Week 2:

Data Engineering Lite & Model Lifecycle

Theme: Engineering Data Before Models.

Core Principle:

Bad Data = Bad Model

Good Data = Reliable System

Week 2 builds real-world data competence —

the #1 failure point in AI systems.

Week 2 —

6 Lesson Structure

Lesson 1:

Understanding Datasets & Data Splits

Focus:

Structured vs Unstructured Data

Features vs Labels

Training / Validation / Test Split

Data Leakage (what it is & why it destroys models)

Dataset documentation mindset

Outcome: Students understand how datasets are structured and how improper splitting ruins evaluation.

Lesson 2:

Pandas for Data Loading & Exploration

Focus:

Loading CSV, JSON, Excel files

Inspecting data (head, info, describe)

Selecting columns

Filtering rows

Basic statistics

Detecting missing values

Outcome: Students can explore and inspect datasets confidently.

Lesson 3:

Data Cleaning & Transformations (Pipelines Mindset)

Focus:

Handling missing values

Dropping vs imputing

Type conversion

Removing duplicates

Creating derived columns

Introduction to transformation pipelines

We carve this out separately to deepen cleaning discipline.

Outcome: Students can systematically clean messy datasets.

Lesson 4:

Handling Messy African Datasets (Real-World Context)

Focus:

Noisy inputs

Multilingual text handling

Encoding categorical variables

Imbalanced datasets

Data validation mindset

This lesson becomes context-specific and practical.

Outcome: Students can manage imperfect real-world data.

Lesson 5:

Feature Engineering Basics

Focus:

Why features matter more than models

Scaling & normalization

Encoding techniques

Basic text feature extraction

Feature importance intuition

Outcome: Students understand how to transform raw data into ML-ready signals.

Lesson 6:

Model Lifecycle & Production Thinking

Focus:

Data collection

Training

Evaluation

Deployment

Monitoring

Iteration

Dataset drift awareness

This is the integration lesson of the week.

Outcome: Students understand the full AI system lifecycle before deep ML.

Week 2 Final Outcome

By the end of Week 2, students can:

✔ Clean and prepare datasets

✔ Build structured data pipelines

✔ Avoid leakage

✔ Handle messy real-world data

✔ Engineer features

✔ Understand the full model lifecycle

✔ Think production-first.

Now Week 2 is 6 lessons structured.

🔷 Week 2 — Lesson 1

Understanding Datasets & Data Splits.