Data Science on
Cloud

Before diving into cloud-based data science solutions, prepare yourself with these 3 things:

• Brush up on the conceptual understanding of various ML Models
• Brush up on your Python programming skills since we'll be using it extensively.
• If you're new to the cloud, familiarize yourself with its basics and what a data scientist needs in this environment.

You'll receive exclusive access to content designed to support your learning journey upon enrollment.

The objective of this week is to understand what MLOps is, why is it important in today’s business environment, and learn how to navigate the ML Lifecycle.

• What is MLOps?
• The requirement for MLOps in the industry
• Business Examples of ML Solutions for products
• Machine Learning Lifecycle
• Challenges in ML Ops
• Looking forward - setting expectations for the journey ahead

Key Skills

• AWS: Walkthrough of different business cases related to machine learning using AWS SageMaker
• Azure: Walkthrough of different business cases related to machine learning using AzureML

The objective of this week is to get set up with the cloud studio you will be working with, the database that you will query for data, and understand how to use the cloud shell.

• What is a warehouse?
• What is the difference b/w a traditional & cloud warehouse

Key Skills 

• AWS: SageMaker Studio, AWS Athena and Glue
• Azure: AzureML Studio, AzureSQL Database

The objective of this week is to learn how to run pre-processing data jobs on the cloud - which can include crucial steps like data quality assessment, cleaning, and transformation.

• Data Quality Assessment
• Data Cleaning
• Data Transformation - Scaling, Encoding, Normalizing
• Outlier detection and handling
• Feature Selection

Key Skills

• AWS: SageMaker Studio, Notebooks, Pre-processing Jobs
• Azure: AzureML, Notebooks, Pre-processing Jobs

The objective of this week is to learn how to build a model, train, tune, and quickly deploy that model as an endpoint for real-time inference.

• How to train and tune the ML model
• General Code Architecture for each stage

Key Skills: 

• AWS: Notebooks, Training Jobs, Hyper-parameter Tuning Jobs, Endpoint Deployment
• Azure: Notebooks, Training Jobs, Hyper-parameter Tuning Jobs, Endpoint Deployment

The objective of this week is to learn how to deploy a model on a container, create an endpoint, and get the inference by invoking the endpoint. 

• Real-time vs Batch Transformation
• Web APIs - how do they work for hosting a model?
• Containerization with docker

Key Skills: 

• AWS: ECS, ECR, EC2
• Azure: Azure Blue Green deployment, Azure Machine Learning service

The objective of this week is to catch up on your practice and attend an optional Q&A session to get help for the mid-term project.

• The Office of Foreign Labor Certification (OFLC) experiences a surge in labor certification applications, resulting in a time-consuming review process. 
• To alleviate this, a machine learning-based solution is needed to streamline candidate shortlisting and increase the likelihood of VISA approval. 
• Enter EasyVisa, the cloud-based, data-driven solution hired by OFLC to automate the visa approval process. As a data scientist at EasyVisa, the role involves analyzing data and using a classification model to automate visa approval.

The objective of this week is to learn the essential concepts of model monitoring, like drift, bias, and explainability and learn how to implement model monitoring on the cloud for your solution.

• What is Model Monitoring?
• Why do you need Model Monitoring?
• Data drift and Model drift
• How to detect drifts?
• Baselining
• Model Monitoring Architecture
• General Code Architecture 

Key Skills: 

• AWS: Amazon CloudWatch
• Azure: Azure Monitoring

The objective of this week is to get a strong foundation of the concepts leading to MLops, like - version control, collaboration, and project templates.

• Basics of Version Control
• Introduction to Git
• Git Concepts - repo, commit, push, pull, branch, merge
• Collaborating with Git
• Best practices for coding
• Environments - Dev & Prod
• General Pipeline Architecture 

Key Skills: 

• AWS: SageMaker Projects, SageMaker Pipelines
• Azure: Azure Pipelines

The objective of this week is to learn how to build an end-to-end ML ops pipeline for your data science solution.

• AWS: CodePipeline, Model Registry, CodeCommit
• Azure: Azure Repos, Model Registry

• Last-minute cancellations hurt hotel revenue as they are caused by various factors like changing plans and flexible cancellation policies. Online bookings, although convenient, make cancellations more unpredictable, posing a challenge for hotel management.
• INN Hotels Group needs an ML solution to predict high cancellation bookings. As an ML Engineer, develop and deploy a model, clean and transform data, and provide a cloud-based API for stakeholders. 
• Goal: Boost INN Hotels' revenue by reducing cancellations through accurate predictions.