Machine
Learning
System Design

Course Overview

How does Netflix create personalized content for each of their users?

How do we get beyond Untitled54.ipython Jupyter Notebook?
How do we build models to predict sales for 1000 items?

The answer is Machine Learning System Design! Organizations in today's world need an automated and streamlined ML process. This process does not just help the organization successfully deploy ML models in production but also optimize and maintain those at scale.

ML System Design is designed to help students transition from classroom learning of machine learning to real world application. It is aimed at the nuances within the industry where data is often messy and there is usually no single model that fits all. Through this course, my aim would be emulate real life examples as much possible to help students

Canvas

Class Github

Modulely Schedule

This is a pre-class session where you will need to sign up on Google Cloud Platform using the Education Grant
Please install:
1. Github Desktop
2. Anaconda

For our premier class, we will begin by discussing the ecosystem of what entails ML System Design and some basics of ML Ops. We will discuss forecasting at scale with emphasis on the following:
- Running multiple time series concurrently
- Hyperparameter Tuning and cost optimization
- Asymmetric loss function
- Interquartile range calculation
- Model Deployment

We will beging the class with a short discussion on functions and classes. We will then practice building our very first algorithm from scratch

Utilizing Transfer Learning
Zero Shot models
Lab: Let's build a contextual search engine!

We will discover the magic of Streamlit and use it to build beautiful apps and deploy them

Walkthrough of Airflow
Final Presentation

Selected Works

Here are some of the projects we will be working on. Currently a placeholder

Cookie Recipe Finder

Project 1

This app will take two user inputs: the ingredients they wish to use in their cookies and the maximum amount of time they have to make their cookies.