Week 8-9 : Supervised Learning: Regression (Days 44-57)
1. Introduction
Weeks 8–9 in a typical ML curriculum cover a familiar arc: linear regression from scratch, scikit-learn pipelines, a capstone housing model, then logistic regression, binary classification, multiclass strategies, and a production spam detector. Days 52–57 often shift to review (linear algebra, probability, statistics). That pause is useful—but the modeling lessons themselves still risk living in eight separate folders, eight scripts, and eight mental models.
Software engineers do not ship “Day 46.” They ship services: train artifacts, expose predict endpoints, compare modalities, and run composed workflows behind one API.
For a limited time, get the Hands-On AI Engineering Course at 50% off.
Build AI-powered apps from scratch — with quizzes, projects & real code.
Join 29,000+ engineers already learning. This deal won’t last.
🔥 Claim 50% Off → aieworks.substack.com/2bf5dcd9
Week 8–9: Supervised Learning — Regression & Classification (week_8_9_aiml_integrated_project) unifies Days 44–51 into one system:
A core Python package with ported lesson modules and JSON-safe API wrappers
A composition engine (
supervised_ml_engine) that wires lessons into product workflowsA FastAPI backend with dual surfaces: learning (per day) and product (composed)
An optional React frontend for regression labs, classification demos, housing predict, spam ops, and pipeline runs
Parent day44/–day51/ folders in the repo stay read-only references. All runtime code lives in the integrated project—same pattern as prior week integrations, applied to supervised learning.
2. System Overview
The project root is week_8_9_aiml_integrated_project/. Runtime is deliberately small:
Key components
Core package — packages/week89_python/src/week89_python/
Eight day modules (day44_linear_regression_theory.py … day51_spam_detection_production.py) plus week89_apis.py for serializable entry points used by HTTP handlers.
API layer — backend/app/
Routing (week89_lab.py), Pydantic schemas, request logging, and to_jsonable() for NumPy/pandas types. Handlers stay thin; they do not reimplement lesson logic.
Composition engine — supervised_ml_engine/
Facades under regression/, classification/, and evaluation/, orchestrated by pipeline.py. The engine imports day modules; day modules never import the engine.
Frontend (optional) — frontend/
Pages call /api/v1/week89/* for teaching flows and /api/v1/supervised/* for composed workflows. Swagger at /docs remains the contract reference.
Dual-layer API design
Pedagogy stays isolated. Product engineers get one vocabulary for housing predict, spam train, and end-to-end demos.
3. Architecture Diagram
For a limited time, get the Hands-On AI Engineering Course at 50% off.
Build AI-powered apps from scratch — with quizzes, projects & real code.
Join 29,000+ engineers already learning. This deal won’t last.
🔥 Claim 50% Off → aieworks.substack.com/2bf5dcd9
