Practical MLOps for Data Scientists & DevOps Engineers

A b o ut t h e c o ur s e
Why this course is different ?
Curated by Industry Experts
We thank you for choosing us!!
Created with the approach to take the practitioner
from Academia to Realtime business use case

Best Practices in Production Usage

Notebooks and Scripts to use it on your client

projects with little modifications

Gain the most sought-after skillset in Data Science

About Instructor
• Murthy is a Data Science Manager working in a Top Consulting Firm and sharing
his knowledge through Manifold AI Learning. His Day-to-Day work includes:
• Leading a Team of 25 People and is responsible for End-to-End Implementation
of Data Science, Machine Learning & Deep Learning Projects
• Manage the End-to-End Pipeline starting from Experimentation to
Operationalization of Machine Learning Models in the Production Environment,
including the maintenance of Machine Learning Models.
• Worked on Various domains like – Retail, Healthcare, Banking & Insurance.

• Impacts Made:
• Successfully Led & delivered multiple Data Science Projects from POC to
Production with overall revenue close to 100m$
• Leader on MLOps Approach in my Organization
• Author of 10 Top selling e-learning content on Data Science
• Trained more than 50k learners through Live Bootcamps
• Top Instructor on Data Science with Manifold AI Learning
Share your ML Projects & Your Experimentation by
Tagging us on LinkedIn
Bootcamp for Limited Learners

Link - https://www.manifoldailearning.in/courses/MLOps---LLMOps-AIOps-Bootcamp---A-
Private-Community-65c63808e4b09cf855132de4
We are Excited to see you on our
class!!
Practical MLOps for Data Scientists & DevOps Engineers
What & Why MLOps ?
What is MLOps ?

MLOps stands for Machine Learning Operations

MLOps is a core function of Machine Learning

engineering, focused on streamlining the
process of taking machine learning models to
production, and then maintaining and
monitoring them

MLOps is a collaborative function, often

comprising data scientists, devops engineers,
and IT
Who are all Key People in a Machine Learning Projects ?

Data Engineer

Data Scientist

ML Engineer

Business Stakeholder

Data Governance Officer

Process of Machine Learning

Data Preparation

Exploratory Data Analysis

Feature Engineering

Model Training

Model Validation

Model Deployment

Monitoring
Why MLOps are the Need Right Now ?

• Building the Models for Personal Project is different than the one
built for Organization
• Take care of Dependencies
• Communication between Multiple Teams
• Managing the Model Risks
• Manage the Infrastructure for Scale
 Solution ?

A Streamlined approach for Machine Learning Projects à MLOps

Some Guiding Principles

• Always Keep Business Goals in mind

• Have a Data Centric Approach
• Modular Approach in Implementation
• Mature in a way such that Process should guide
automation
Next Video :

Fundamentals of MLOps
Goal of MLOps on Use Case

Take care of End To End Process of Data Science

from Data Preprocessing to Deployment
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

MLOps Fundamentals
Environments & Assets in Machine Learning
Projects
Environments

1. Dev – can be accessed by people –no guarantee of quality – low risk

2. Stage - can be accessed by subset of people – test of dev code –
quality is matched with prod – treated as a live product
3. Prod – Very least number of people has access (tight control on
access) – Live Product –highest quality in assets – business critical
Assets

• ML Project Code – Stored in Version Control Repository (Git)

• Models – Model & Model Artifacts with Model Registry
• Data – Separate environments for storage.
• Dev – Temporary
• Stage – Almost same as Prod – updates on frequent basis (Monthly, Quarterly, etc.)
• Prod – Reliable and Fresh
CI CD Pipeline
Next Video :

Deep Dive into MLOps

MLOps Deep Dive
Understanding Process
CI CD Pipeline
MLOps

Data : Read-only access to production data and read-write

access to a separate dev storage environment to develop
and experiment with new features and other data tables.

EDA : Exploration of data, prepare the data for the Applying

on model
MLOps – Contd.

Repository : A Place to store all the source code with version

enabled

Feature Table : Reading the raw data table and feature tables
and writes to tables in Feature Store.
MLOps – Contd.

Model Training : Create Model Training Pipeline, involving –

Training & Tuning, Evaluation and Model Output

Commit the Repo: Update the Dev Branch with updated code
MLOps – Contd.

Dev Environment: Perform Integration Test

Merge the Code : Merge with Stage Branch

MLOps – Contd.

Stage Environment: Perform Integration Test (CI)

Merge the Code : Run Integration testing with latest data

Deployment

Deployment Training : Create Model Training Pipeline,

involving – Training & Tuning, Evaluation and Model Output

Commit the Repo: Update the Dev Branch with updated code
Continuous Deployment

• Compliance Checks
• Comparison between Stage Vs Prod
• Deploy Model to Prod
Deployment Strategies

• Online Serving (REST APIs)

• Batch/Streaming inference
Monitoring

• Data Ingestion
• Accuracy & Data Drift
• Monitor Metrics
• Retrigger Model Training
Retrain of Models

• Scheduled
• Trigger based on Events
WHY DEVOPS IS
NOT SAME AS
MLOPS
 • Although DevOps and MLOps are well-
known terms among data and analytics
professionals, many business executives
do not fully grasp the magnitude of the
Machine Learning revolution sweeping
across industries, nor do they understand
the significance of these terms for their
DEVOPS VS organizations in effectively managing ML.
MLOPS • As a result, many organizations struggle
to operationalize their ML efforts because
they mistakenly apply their well-
established DevOps principles to ML
without recognizing the differences and
the role these practices play within a
business strategy.
QUICK RECAP ON
DEVOPS

• In recent decades, DevOps has gained

significant popularity as an approach to
effectively managing software development.
• The term "DevOps" was introduced by Patrick
Debois, a renowned software developer and
consultant, in 2009.
• The origins of DevOps can be traced back to
the early 2000s, when companies started
realizing the importance of enhanced
collaboration and communication between
software developers and IT operations teams.
WHY DEVOPS IS NOT
SUITABLE FOR MACHINE
LEARNING

• While DevOps may be suitable for managing

conventional software development, it is not the
most optimal approach for the development and
operationalization of ML models.

• ML models possess the capability to learn from

data and make predictions or classifications
without explicit programming, leading to
significant advancements in fields like image
recognition, natural language processing, and
predictive analytics.

• The development and deployment of ML models

necessitate specialized infrastructure, efficient
data management, robust processing capabilities,
and expertise in specialized skills and knowledge
to ensure effective implementation.
UNIQUE CHALLENGES OF
MACHINE LEARNING

• ML development is characterized by its iterative and

experimental nature, distinguishing it from traditional software
development practices. and knowledge: ML requires
specialized skills and knowledge in areas such as data
engineering, data science, statistics, and math - and how to
apply these in the context of a business use case

• ML necessitates specialized skills and expertise in areas such as

data engineering, data science, statistics, and mathematics,
and their application within specific business use cases.

• ML models often demand specialized infrastructure and

resources, including GPUs, TPUs, distributed computing
systems, and tools for data storage and management.

• Once a model is developed, it must be effectively deployed and

managed within a production environment.
THE UNIQUE CHALLENGES
OF MACHINE LEARNING

• Effective data management and

governance are crucial for the success
of ML development because ML
models heavily depend on extensive
data for training and making accurate
predictions. Therefore, ensuring
proper data management and
governance becomes paramount in
the ML development process.
 FINAL

DevOps - this is the way for Traditional Software development

MLOps - this is the highway - Needs more process oriented

approach to overcome the challenges
THANK YOU
W H AT & W H Y
AWS
OVERVIEW OF AMAZON
WEB SERVICES

• Amazon Web Services offers a broad set of

global cloud-based products including
compute, storage, databases, analytics,
networking, mobile, developer tools,
management tools, IoT, security, and
enterprise applications: on-demand, available
in seconds, with pay-as-you-go pricing
• This allows enterprises, start-ups, small and
medium-sized businesses, and customers in
the public sector to access the building blocks
they need to respond quickly to changing
business requirements
INTRODUCTION TO
AWS

• In 2006, Amazon Web Services began offering

IT infrastructure services to businesses as web
services—now commonly known as cloud
computing
• One of the key benefits of cloud computing is
the opportunity to replace upfront capital
infrastructure expenses with low variable costs
that scale with your business
• With the cloud, businesses no longer need to
plan for and procure servers and other IT
infrastructure weeks or months in advance
 • Cloud computing is the on-demand
delivery of compute power, database,
storage, applications, and other IT
resources through a cloud services
platform via the internet with pay-as-you-
go pricing

• Cloud computing provides a simple way

W H AT I S C LO U D to access servers, storage, databases and
a broad set of application services over
COMPUTING? the internet

• A cloud services platform such as

Amazon Web Services owns and
maintains the network-connected
hardware required for these application
services, while you provision and use
what you need via a web application
S I X A D VA N TAG E S O F
CLOUD COMPUTING

• Trade fixed expense for variable expense – Instead of

having to invest heavily in data centers and servers
before you know how you’re going to use them, you
can pay only when you consume computing
resources, and pay only for how much you consume

• Benefit from massive economies of scale – By using

cloud computing, you can achieve a lower variable
cost than you can get on your own

• Stop guessing capacity – Eliminate guessing on your

infrastructure capacity needs

• Increase speed and agility – In a cloud computing

environment, new IT resources are only a click away,
which means that you reduce the time to make those
resources available to your developers from weeks to
just minutes
SIX ADVANTAGES OF
CLOUD COMPUTING
CONTD.

• Stop spending money running and

maintaining data centers – Focus on
projects that differentiate your
business, not the infrastructure

• Go global in minutes – Easily deploy

your application in multiple regions
around the world with just a few clicks
TYPES OF CLOUD
COMPUTING

• Cloud computing provides developers and IT

departments with the ability to focus on what
matters most and avoid undifferentiated work
such as procurement, maintenance, and capacity
planning
• As cloud computing has grown in popularity,
several different models and deployment
strategies have emerged to help meet specific
needs of different users
• Understanding the differences between
Infrastructure as a Service, Platform as a Service,
and Software as a Service, as well as what
deployment strategies you can use, can help you
decide what set of services is right for your needs
 CLOUD
COMPUTING
MODELS
INFRASTRUCTURE
AS A SERVICE

• Infrastructure as a Service contains the

basic building blocks for cloud IT and
typically provides access to networking
features, computers , and data storage
space
• IaaS provides you with the highest level
of flexibility and management control
over your IT resources and is most
similar to existing IT resources that
many IT departments and developers
are familiar with today
P L AT FO R M A S A
SERVICE

• Platform as a Service removes the

need for your organization to manage
the underlying infrastructure and
allows you to focus on the deployment
and management of your applications
• This helps you be more efficient as you
don’t need to worry about resource
procurement, capacity planning,
software maintenance, patching, or
any of the other undifferentiated heavy
lifting involved in running your
application
SOFTWARE AS A
SERVICE

• Software as a Service provides you with a

completed product that is run and managed by
the service provider

• With a SaaS offering you do not have to think

about how the service is maintained or how the
underlying infrastructure is managed; you only
need to think about how you will use that
particular piece of software

• A common example of a SaaS application is web-

based email which you can use to send and
receive email without having to manage feature
additions to the email product or maintain the
servers and operating systems that the email
program is running on
CLOUD COMPUTING DEPLOYMENT
MODELS
CLOUD
HYBRID

• A hybrid deployment is a way to connect

infrastructure and applications between cloud-
based resources and existing resources that
are not located in the cloud
• The most common method of hybrid
deployment is between the cloud and existing
on-premises infrastructure to extend, and
grow, an organization's infrastructure into the
cloud while connecting cloud resources to the
internal system
• For more information on how AWS can help
you with your hybrid deployment, visit
our Hybrid Cloud with AWS page
ON-PREMISES

• The deployment of resources on-premises,

using virtualization and resource management
tools, is sometimes called the “private cloud.”
• On-premises deployment doesn’t provide
many of the benefits of cloud computing but is
sometimes sought for its ability to provide
dedicated resources
• In most cases this deployment model is the
same as legacy IT infrastructure while using
application management and virtualization
technologies to try and increase resource
utilization
GLOBAL INFRASTRUCTURE OF AWS
GLOBAL
INFRASTRUCTURE

• The AWS Cloud infrastructure is built around

AWS Regions and Availability Zones
• Availability Zones consist of one or more
discrete data centers, each with redundant
power, networking, and connectivity, housed in
separate facilities
• These Availability Zones offer you the ability to
operate production applications and
databases that are more highly available, fault
tolerant, and scalable than would be possible
from a single data center
BENEFITS OF AWS
SECURITY

• Keep Your data safe — The AWS

infrastructure puts strong safeguards
in place to help protect your privacy

• Meet compliance requirements —

AWS manages dozens of compliance
programs in its infrastructure

• Save money: —Cut costs by using

AWS data centers

• Scale quickly — Security scales with

your AWS Cloud usage
 Technical Stack of AWS for MLOps &
NEXT VIDEO Machine Learning
AWS
Technical Stack of AWS for MLOps & Machine
Learning
ML projects are multi-disciplinary and iterative
Data
Business Curation
Understanding Data
Access Feature
Data
Collection
Exploratory Engineering People
Analysis
Data Profiling and
Cleaning Algorithm Debugging
Building
Model Experiment
Training Parameter Tracking
Tuning
Governance
& Approvals
Bias & Process Tools
Model Explainability
Risk Deployment
Online
Management Environment Monitoring
Systems Management Business
Integration Metrics …In other words, “difficult”!
2
Amazon SageMaker and the AWS AI/ML Stack
SPECIALIZED

Business processes Search Code + DevOps Industrial Healthcare

Amazon Personalize Amazon Kendra Amazon CodeGuru Amazon Monitron Amazon HealthLake
Amazon Forecast Amazon DevOps Guru Amazon Lookout for Equipment Amazon Comprehend Medical
Amazon Fraud Detector Amazon Lookout for Vision Amazon Transcribe Medical
Amazon Lookout for Metrics
AI SERVICES
CORE

Text & Documents Chatbots Speech Vision

Amazon Translate Amazon Lex Amazon Polly Amazon Rekognition
Amazon Comprehend Amazon Transcribe AWS Panorama
Amazon Textract Amazon Transcribe Call Analytics

SAGEMAKER SAGEMAKER
CANVAS STUDIO LAB
AMAZON SAGEMAKER

ML SERVICES Label No-code ML Learn Prepare Store Detect Build with Train Tune Deploy in Explain Manage Manage
data for business ML data features bias notebooks models parameters production predictions & monitor edge devices
analysts CI/CD

ML FRAMEWORKS PyTorch, Apache Habana Elastic

Amazon EC2 CPUs GPUs AWS Inferentia AWS Trainium FPGA
& INFRASTRUCTURE MXNet, TensorFlow Gaudi inference
Modular features for the end-to-end ML lifecycle
PREPARE BUILD TRAIN & TUNE DEPLOY & MANAGE
Geospatial Studio Notebooks & Notebook Instances Fully Managed Training Fully Managed Deployment
Visualize geospatial data Fully managed Jupyter notebooks with elastic compute Broad hardware options, easy to setup and scale Ultra low latency, high throughput inference

Ground Truth Studio Lab Automatic Model Tuning Real-Time & Serverless Inference
Free ML development environment For steady and intermittent traffic patterns
Create high quality datasets for ML Hyperparameter optimization
Autopilot Asynchronous Inference
Data Wrangler Automatically create ML models with full visibility Customization Support For large payloads or long processing times
Aggregate and prepare data for ML Integrate with popular open source frameworks
Batch Transform
Built-in Algorithms and libraries
For offline inference on batches of large datasets
Processing Integrated tabular, NLP, and vision algorithms
Built-in Python, BYO R/Spark Experiments Multi-Model Endpoints
Bring Your Own Track, visualize, and share model artifacts Reduce cost by hosting multiple models per instance
Bring your own container and algorithms across teams
Feature Store Multi-Container Endpoints
Store, catalog, search, and reuse features Reduce cost by hosting multiple containers
JumpStart Managed Spot Training per instance
UI based discovery, training, and deployment of
Clarify Reduce training cost by up to 9 0 %
models, solutions, and examples Shadow Testing
Detect bias and understand model predictions Validate model performance in production
Local Mode Debugger and Profiler
Test and prototype on your local machine Debug and profile training runs Inference Recommender
Automatically select compute instance
and configuration
Distributed Training Libraries
High performance training for large datasets Model Monitor
and models Maintain accuracy of deployed models

Kubernetes Operators & Components

Training Compiler Manage and monitor models on edge devices
Faster deep learning model training
Edge Manager
Manage and monitor models on edge devices

MLOps: Pipelines | Projects | Model Registry Canvas Studio | RStudio Governance

Workflow automation, CI/CD for ML, Generate accurate machine learning Integrated development Model Cards | Dashboard |
central model catalog predictions—no code required environment (IDE) for ML Permissions

4
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

MLOps
How to make the most of the course !!
Key Recommendations

Recommended to learn in sequence

Feel free to Jump into specific section to make the most

for your use-case

Once a module is completed, recommended to review and

think about its learning & its implementation for your
personal & client work

Keep in mind – Some Services are chargeable and not part

of Free-tier on AWS – Ensure to stop services once the
learning activity is completed

Practice on your personal project (No Alternate for this)

 Understanding Software
Development Life Cycle: Its
Benefits & Challenges
Introduction to SDLC
 Phases Of SDLC

• Planning • Release Phase

Identifying the scope of the project and setting goals Team Packaging, managing & deploying releases across
and objectives. various environments.

• Coding Phase • Deployment

System Design, static code analysis and code review. Deploying the application to production environment.

• Building Phase • Operate

Building the Software Use of Software in the production Environment

• Testing • Monitor
Testing the application to ensure it meets requirements. Software, System, User Experience & Security are
monitored
 Processes Involved in SDLC

Phase 7:
Phase 1: Planning Phase 3: Design Phase 5: Testing Maintenance
Define project scope, Develop a detailed Test the software to Provide ongoing
objectives, and design of the system. ensure it meets support and
timeline. requirements. maintenance for t
software.

Phase 2: Analysis Phase 4: Phase 6:

Analyze user Development Deployment
requirements and create Develop the software Deploy the software to
a system design. according to the production
design. environment.
The Benefits
 The Advantages of Using the SDLC

Improved Quality Reduced Risk Increased Efficiency

SDLC helps to ensure that the product meets SDLC helps to identify and mitigate risks SDLC helps to streamline the development
customer requirements and is of high quality. early in the development process. process, resulting in faster delivery times.

The use of SDLC can help organizations to improve the quality of

their products, reduce risk, and increase efficiency.
Next Video :
Waterfall Vs
Agile Vs
DevOps
In the face of changing times, organizations
are prompted to reassess their current
production methodologies.

While each organization has its own distinctive

approach to work, they typically rely on three
highly effective methodologies: Waterfall, SDLC
Agile, and DevOps.

These three methodologies have been

thoroughly tested and proven to deliver
favorable outcomes for organizations
throughout their existence.
Comparing
Waterfall vs
Agile vs Devops
Waterfall
Waterfall Approach
Advantages of Waterfall

The presence of a partially established framework enables the

subsequent team to make modifications or corrections to the
project.

This sequential process ensures that the client receives a product

that is the result of collaborative team efforts.

What gives Waterfall an advantage over the other two

methodologies is that it does not require any certification for
individuals to participate in the Waterfall production.
 This approach can lead to delays in
development and require the
Waterfall operates as a production line development team to invest additional
for software development. resources in exploring alternative
methods, as Waterfall restricts changes

Cons of
once a stage is completed.

It is crucial to establish the end goals

Waterfall
from the beginning, as this enables
Favored for Small scale projects
effective planning of the various stages
involved.
Agile
`
Agile Methodology

Agile methodology is a flexible and iterative approach to software

development and project management. It emphasizes adaptability,
collaboration, and continuous improvement throughout the
development process.

In agile, the development process is divided into short iterations called

sprints, typically lasting one to four weeks.
 Principles of Agile Methodology

CUSTOMER ITERATIVE ADAPTIVE SELF-ORGANIZING CONTINUOUS

COLLABORATION DEVELOPMENT PLANNING TEAMS IMPROVEMENT
Advantages of
Agile
• W h i l e t h e Wa t e r f a l l a p p r o a c h y i e l d s r e l i a b l e
outcomes, Agile methodology alleviates the
burden on the production team by facilitating
error handling and failure mitigation through
enhanced collaboration.

• The adoption of Agile leads to a considerable

r e d u c t i o n i n d e v e l o p m e n t- t o - d e p l o y m e n t t i m e a s
the development team can focus on releasing
completed components.

• C o n s e q u e n t l y, t h e t e s t i n g t e a m n o l o n g e r n e e d s
t o w a i t f o r t h e e n t i r e p a c k a g e t o b e a va i l a b l e
before conducting bug testing.
Software developed using Agile methodology does
not have a fixed endpoint or final version.

As long as there is demand, the company is

obligated to continuously release updates and
improvements.

Apple, for example, has consistently introduced new

versions of iOS since the initial release of the
Cons of
iPhone.
Agile
Organizations must be prepared to address this
ongoing demand.

While this is a notable advantage of Agile, it can

also be a disadvantage as each team, working
within their own sprint, develops their own set of
deliverables.
 DevOps

DevOps stands apart from its counterparts by amalgamating the most

valuable aspects of the Agile process with IT production teams.

In contrast to the fragmented nature of working independently, where

miscommunication and delayed rollouts were prevalent due to the need for
coordination among numerous team members, DevOps provides a
revolutionary solution by merging these teams together.
 AGILE VS DEVOPS

While Agile methodology focused on flexibility, enabling team members to

rectify mistakes, DevOps emphasizes automation to save time and effort.

Since DevOps is derived from Agile, the principle of completing work in

iterations remains. By adopting this approach, the pressure on DevOps
team members is alleviated, as they can develop, test, and deploy
components in a more coordinated manner through incremental updates.
Next Video
DevOps Lifecycle and
Tools in AWS
DEVOPS LIFECYCLE &
TOOLS IN AWS
 • DevOps is the combination of
cultural philosophies, practices,
and tools that increases an
organization’s ability to deliver
applications and services at high
velocity: evolving and improving
products at a faster pace than
organizations using traditional
software development and
infrastructure management
processes
• This speed enables organizations
to better serve their customers
and compete more effectively in
the market

DEVOPS MODEL
 • Under a DevOps model,
development and operations
teams are no longer “siloed.”
• In some DevOps models, quality
assurance and security teams
may also become more tightly
integrated with development and
operations and throughout the
application lifecycle
• These tools also help engineers
independently accomplish tasks
that normally would have
required help from other teams,
and this further increases a
team’s velocity

HOW DEVOPS
WORKS
Source : AWS
DEVOPS • Continuous
Integration
PRACTICES
• Continuous Delivery
• Microservices
• Infrastructure as
Code
• Monitoring and
Logging
• Communication and
Collaboration

Source : AWS
 • Continuous integration is a
software development practice
where developers regularly
merge their code changes into a
central repository, after which
automated builds and tests are
run
• The key goals of continuous
integration are to find and
address bugs quicker, improve
software quality, and reduce the
time it takes to validate and
release new software updates

CONTINUOUS
INTEGRATION
Source : AWS
 • Continuous delivery is a
software development practice
where code changes are
automatically built, tested, and
prepared for a release to
production
• It expands upon continuous
integration by deploying all code
changes to a testing
environment and/or a production
environment after the build stage
• When continuous delivery is
implemented properly,
developers will always have a
deployment-ready build artifact

CONTINUOUS that has passed through a

standardized test process

DELIVERY
Source : AWS
 • The microservices architecture
is a design approach to build a
single application as a set of
small services
• Each service runs in its own
process and communicates with
other services through a well-
defined interface using a
lightweight mechanism, typically
an HTTP-based application
programming interface
• Microservices are built around
business capabilities; each
service is scoped to a single
purpose

MICROSERVICES
Source : AWS
 • Infrastructure as code is a practice in which
infrastructure is provisioned and managed using
code and software development techniques, such
as version control and continuous integration
• The cloud’s API-driven model enables developers
and system administrators to interact with
infrastructure programmatically, and at scale,
instead of needing to manually set up and
configure resources
• Thus, engineers can interface with infrastructure
using code-based tools and treat infrastructure in
a manner similar to how they treat application
code
INFRASTRUCTURE
AS CODE

Source : AWS
 • Developers and system administrators use
code to automate operating system and
host configuration, operational tasks, and
more
• It frees developers and systems
administrators from manually configuring
operating systems, system applications, or
server software

CONFIGURATION
MANAGEMENT

Source : AWS
 • With infrastructure and its configuration
codified with the cloud, organizations can
monitor and enforce compliance
dynamically and at scale
• This makes it easier for organizations to
govern changes over resources and
ensure that security measures are
properly enforced in a distributed manner
• This allows teams within an organization
to move at higher velocity since non-
compliant resources can be automatically
flagged for further investigation or even
automatically brought back into
POLICY AS CODE compliance

Source : AWS
 • Organizations monitor metrics and logs to see
how application and infrastructure performance
impacts the experience of their product’s end user
• By capturing, categorizing, and then analyzing data
and logs generated by applications and
infrastructure, organizations understand how
changes or updates impact users, shedding
insights into the root causes of problems or
unexpected changes
• Active monitoring becomes increasingly important
as services must be available 24/7 and as
application and infrastructure update frequency
increases
MONITORING
AND LOGGING

Source : AWS
 • Increased communication and
collaboration in an organization
is one of the key cultural aspects
of DevOps
• The use of DevOps tooling and
automation of the software
delivery process establishes
collaboration by physically
bringing together the workflows
and responsibilities of
development and operations
• AWS provides a set of flexible
services designed to enable
COMMUNICATION companies to more rapidly and
reliably build and deliver

AND products using AWS and DevOps

practices

COLLABORATION
Source : AWS
AWS TOOLS FOR DEVOPS

Source : AWS
NEXT VIDEO

Getting Started with Amazon Web Services – For Beginners

14
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Amazon Web Services

Getting Familiar with Basics of AWS
Agenda of the Section

- Creation of AWS Account

- Setting Up of Account (IAM User and
MFA)
- Configure CLI
- Hands On – IAM
- IAM Policy Generator
- S3 Bucket (Creation, Versioning,
Lifecycle Management)
- EC2 Instances (Launch & SSH)
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Amazon Web Services

S3 – Simple Storage Service
Agenda
What is S3 Bucket?

S3 Storage Class

Types of S3 Storage Classes

MANIFOLD AI LEARNING®
What is S3
Bucket? Amazon Simple Storage Service is a storage
service that can be maintained and
accessed over the Internet
Using the web service, we can store and
retrieve an unlimited amount of data

S3 in Amazon has two entities called

buckets and objects

MANIFOLD AI LEARNING®
AWS S3 Bucket
Benefits 99.99% durability

5GB of Amazon S3 standard storage

S3 provides Encryption to store

Multiple copies are maintained to enable the restoring the data in case of
data corruption

We can S3 is Highly Scalable since it automatically scales your storage

according to your requirement

Pay per use Model

MANIFOLD AI LEARNING®
Amazon S3
Storage Classes
Amazon S3 offers a wide range of storage classes for
various use cases

Gives power to the user for the storage of data

appropriately as per the user’s need.

All Amazon S3 storage classes have a high level of reliability

and support SSL data encryption during transmission, only
cost is different depending on storage classes

MANIFOLD AI LEARNING®
Types of S3 Bucket
Storage Classes

S3 Standard
S3 Standard-IA
S3 Intelligent-Tiering
S3 One Zone-IA
S3 Glacier
S3 Glacier Deep Archive
S3 Outposts

MANIFOLD AI LEARNING®
S3 Standard
High Availability and low latency

Data is stored in multiple locations

The durability of 99.999999999% and availability

of 99.99% availability over a given year

Most expensive storage class among all others

MANIFOLD AI LEARNING®
S3 Intelligent-
Tiering Low latency and high throughput performance

Automatically moves the data between two access

tiers. (Infrequent Access and Frequent Access)

The durability of 99.999999999% and availability

of 99.99% availability over a given year

Small monthly monitoring and auto-tiering fee

MANIFOLD AI LEARNING®
S3 One Zone-IA

Low Latency and High throughput

performance
The durability of 99.999999999% and
availability of 99.5% availability over a
given year
Data will be lost if the Availability Zone
where the data is stored is destroyed
Suitable for larger objects greater than
128 KB kept for at least 30 days

MANIFOLD AI LEARNING®
S3 Glacier

Low-cost design for long-term archiving

Data will be available in case of entire Availability

Zone destruction

The durability of 99.999999999% and availability

of 99.9% availability over a given year

It has a minimum storage duration period of 90

days

MANIFOLD AI LEARNING®
S3 Glacier Deep Archive

The durability of
Lowest cost storage option 99.999999999% and
in S3 availability of 99.9%
availability over a given year

Retrieval costs can be

It has a minimum storage
reduced by using bulk
duration period of 180 days
retrieval

MANIFOLD AI LEARNING®
S3 Outposts

provides object storage easy to store, retrieve, S3 Object compatibility

For durable and
to our on-premises secure, control access, and bucket
redundant storage of
AWS outposts tag, and report on the management is through
data on Outposts
environment. data. S3 SDK

S3 on Outposts will give

users 48TB or 96TB of
S3 storage capacity,
with up 100 buckets on
each Outpost

MANIFOLD AI LEARNING®
S3 Standard-IA
S3 Standard-Infrequent Access is optimized for
long-lived and less frequently accessed data but
requires rapid access whenever required

Similar to S3 Standard, it also offers high durability,

low latency, and high throughput but has a low per
GB storage price and per GB retrieval fee

This is ideal for backups, long-term storage, and as

a data store for disaster recovery

MANIFOLD AI LEARNING®
Next Video

S3 BUCKET HANDS ON

MANIFOLD AI LEARNING®
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Linux Operating System

DevOps & Data Scientists
Agenda of this Section:

This section is intended to help the

practitioners to become aware of the
Linux Operating System which is a
basic for working on Command Line
Interface on Servers
Infrastructure used in Hands-On Videos

EC2 Instance with Amazon Linux 2 AMI

Next Video:

Basics of Linux
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Basics of Linux
DevOps & Data Scientists
Agenda
What is Linux

Components of Linux

Why Linux ?

Linux Distro

Bash in Linux

Package Manager in Linux

What is Linux ?
Linux is an operating system similar to
Windows, iOS, and Mac OS.
It is worth noting that Android, a highly
popular platform worldwide, operates on
the Linux operating system.
An operating system is a software that
oversees the hardware resources of your
computer. In essence, it facilitates the
communication between your software
and hardware components.
Without the operating system (OS), the
software would be unable to operate.
 Components of Linux
Operating System:
• Bootloader – This software manages the computer's boot process
• Kernel – The kernel, known as "Linux," is the core component of
the system
• Init system – This subsystem initiates the user space and controls
daemons
• Daemons – These are background services like printing, sound, and
scheduling that either start during boot or after logging into the
desktop
• Graphical server – Also referred to as the X server or X, this
subsystem handles graphics display on the monitor
• Desktop environment – This is the part of the operating system
that users directly interact with
• Applications – While desktop environments have limited
applications, Linux, like Windows and macOS, provides a vast
selection of high-quality software titles
Why use Linux?

Open-source nature: Linux is an open-source operating system, which means its

source code is freely available to the public.

Stability and reliability: Linux is known for its stability and reliability.

Security: Linux has a strong focus on security. Due to its open-source nature,
security vulnerabilities can be quickly identified and fixed by the community.
Why use Linux? (Contd.)

Flexibility and customization: Users have the freedom to choose from various desktop
environments, customize their system's appearance, and configure it to their preferences.

Software and package management: Linux distributions provide centralized software

repositories and package management systems.

Community and support: Linux has a large and vibrant community of users and developers
worldwide.
Why use Linux? (Contd)

Cost-effective: Linux is
often considered a cost-
effective option
compared to proprietary
operating systems like
Windows or macOS.
What is a “distribution?”

A Linux distribution, also known as a distro, is a complete

operating system package that consists of the Linux kernel,
various software applications, libraries, and configuration files
Different distributions may have different default software,
desktop environments, package management systems, and
overall design philosophies
They are created and maintained by different organizations or
communities, each with its own goals and target audience
Popular Linux Distributions
Ubuntu: One of the most widely used and beginner-friendly distributions. It
focuses on ease of use, stability, and a large user community.
Debian: Known for its stability, reliability, and adherence to open-source
principles. Debian serves as the base for several other distributions,
including Ubuntu.
Fedora: Developed by the Fedora Project and sponsored by Red Hat. It
emphasizes the use of cutting-edge technologies and serves as a testing
ground for new features that may eventually make their way into Red Hat
Enterprise Linux (RHEL).
CentOS: Based on the source code of Red Hat Enterprise Linux, CentOS aims
to provide a free, community-supported alternative to RHEL. It is popular for
server deployments.
Arch Linux: Designed for advanced users who prefer a DIY (do-it-yourself)
approach. Arch Linux focuses on simplicity, minimalism, and user-centric
customization.
openSUSE: A community-driven distribution known for its stability, user-
friendly configuration tools, and professional-grade features
Linux Mint: Built on top of Ubuntu, Linux Mint provides a polished and user-
friendly desktop environment, making it an attractive choice for beginners
transitioning from Windows.
Manjaro: Based on Arch Linux, Manjaro is known for its user-friendly
approach and pre-installed software packages. It offers a balance between
cutting-edge software and stability.
Bash in Linux

Bash, short for "Bourne Again Shell," is a command-line

interpreter and scripting language that is widely used in Linux
and Unix-based operating systems
Bash provides a textual interface for users to interact with the
operating system by executing commands
Bash offers powerful features and capabilities, including
command-line completion, command history, variables,
conditionals, loops, functions, and input/output redirection
Package Managers
in Linux
Package managers in Linux are software tools that handle the
installation, management, and removal of software packages
on a Linux distribution
Package managers simplify the process of installing and
updating software, making it more efficient and convenient for
users
Package managers simplify the software management process
in Linux, making it convenient for users to keep their systems
up to date and install new software with ease
Common Package Managers in Linux

APT

YUM

DNF

Pacman

Zypper
 VERSION
CONTROL SYSTEM
GIT
ENVIRONMENTS FOR APPLICATION

Dev QA INT PROD

CODE DEPLOYMENT….

Dev QA INT PROD

V1 V1 V1 V1
V2 V2 V2 V2
ISSUES…..WITH V2
ROLLBACK ????
 WITHOUT VERSION CONTROL SYSTEM

vOnce Saved, All the changes are permanent and can’t revert.
vNo tracking of information
vRelease Failures – need to redo
vDowntime of Application
WITH VCS

Time Travel and bring back the

working version of code. J
 WHAT IS Version control, also known as source control, is the
VERSION practice of tracking and managing changes to
software code.
CONTROL?
WHAT IS VERSION CONTROL SYSTEMS ?
Version control systems are software tools
that help a software team manage
changes to source code over time.

Version control software keeps track of

every modification to the code in a
special kind of database. If a mistake is
made, developers can turn back the clock
and compare earlier versions of the code
to help fix the mistake while minimizing
disruption to all team members.
VCS ON LOCAL SYSTEM
à The practice of having the CODE saved
in the local PC

à Local DB keeps a record of the changes.

à SCCS (Source Code Control System) & RCS (Revision Control

System)
CHALLENGES!!

Multiple developers cannot work on

same project.

Solution :
Centralized VCS
 CENTRALIZED VCS
Ø In CVC, A Central Repo is maintained where all the versioned files are kept.

Ø Developers can checkout and check-in changes from their own computers.
CHALLENGES!!

If server becomes un-responsive, then entire project is

lost/ cannot proceed

Solution:
Distributed Version Control
DISTRIBUTED VCS
q Version database (CODE) is stored
on every Developer’s local system
and remote server

q Changes are made is Local and then

PUSH to server

q Highly Available
GIT
Git is a mature, actively maintained
open-source project originally
developed in 2005 by Linus Torvalds,
the famous creator of the Linux
operating system kernel.

Having a distributed architecture, Git is

an example of a DVCS (hence
Distributed Version Control System).
AWS CODE COMMIT
AWS CodeCommit is a secure, highly scalable, managed
source control service that hosts private Git repositories.
It makes it easy for teams to securely collaborate on code with
contributions encrypted in transit and at rest.
CodeCommit eliminates the need for you to manage your own
source control system or worry about scaling its infrastructure.
We can use CodeCommit to store anything from code to
binaries.
Supports the standard functionality of Git, so it works
seamlessly with your existing Git-based tools.
GETTING THE SYSTEM READY
git --version à check if its already installed
apt-get update (ubuntu) or yum update (rhel , centos)
apt-get install git or yum install git

https://git-scm.com/downloads
Docker Tutorial
Introduction to
Docker
• Docker is an open-source containerization platform that allows you to package
applications and their dependencies into lightweight, portable containers
• Containerization technology enables applications to run consistently across different
environments, from development to production, without worrying about
differences in operating systems or dependencies
• Docker provides an efficient and scalable solution for deploying and managing
applications, making it a popular choice among developers and DevOps teams
Introduction to
Docker
• Some key benefits of Docker include
• Efficiency: Docker containers are lightweight and share the host operating system, reducing resource usage and enabling
faster deployment times
• Consistency: Containers encapsulate the application and its dependencies, ensuring consistent behavior across different
environments
• Isolation: Each container operates in its own isolated environment, providing security and preventing conflicts between
applications
• Docker has gained widespread adoption in the industry and has a vibrant ecosystem with a large community of contributors and
ready-to-use images
Places where Docker
can be Used ?
• Application Deployment: Docker is commonly used to package and deploy applications
• Microservices Architecture: Docker is well-suited for building and deploying microservices-based
architectures
• Continuous Integration and Continuous Deployment : Docker plays a crucial role in CI/CD pipelines
• Development Environments: Docker provides a consistent and reproducible environment for
developers
• Scalable Infrastructure: Docker can be used to build and manage scalable infrastructure
• Testing and QA: Docker simplifies the process of testing applications across different environments
What is Containerization
Technology

• Containerization technology refers to the practice of encapsulating an application

and its dependencies into a self-contained unit called a container
• Containerization technology enables the bundling of an application with its
dependencies, libraries, and runtime environment, ensuring that it can be executed
reliably on any system that supports containers, regardless of the underlying
infrastructure or operating system
• Containers leverage operating system-level virtualization, where multiple containers
share the same host operating system kernel but are isolated from one another
Docker
Architecture
• The Docker architecture is composed of several components working
together to provide a containerization platform
• Understanding the Docker architecture is essential for effectively
utilizing Docker
• It provides a consistent and portable environment for applications,
simplifies deployment workflows, and enhances scalability and resource
efficiency
Host OS with
Docker Engine
• The host operating system is the underlying operating system on which Docker runs
• Docker Engine, also known as the Docker daemon, is the core component of Docker that interacts with the
host OS to manage containers
• Docker Engine is responsible for container runtime and provides tools for building, running, and managing
containers
• Compatibility: Docker Engine is designed to work on various operating systems, including Linux, Windows, and
macOS
• Linux-based Host OS: Docker originated on Linux and has native support for Linux containers
• Windows and macOS Host OS: On non-Linux systems like Windows and macOS, Docker relies on virtualization
technologies to provide containerization
Host OS with
Docker Engine
• Container Runtime: Docker Engine manages the container lifecycle,
including container creation, starting, stopping, and deletion
• Interaction with Docker CLI: Docker CLI serves as the interface for users
to interact with Docker Engine
• Interacting with the Host OS: Containers run on top of the host OS,
leveraging the underlying kernel for system resources
Docker CLI for
Managing Containers
• The Docker Command-Line Interface is a powerful tool for interacting with Docker Engine and managing containers

• Docker CLI provides a wide range of commands to build, run, manage, and troubleshoot containers effectively

• docker run
• Creates and runs a new container from an image
• Example: docker run -d --name my-container my-image

• docker ps
• Lists running containers
• Example: docker ps
 Stops, starts, or restarts a
container
docker stop/start/restart Example: docker stop my-
container

Docker CLI
for docker rm
Removes one or more
containers

Managing Example: docker rm my-

container

Containers
Displays the logs of a container
docker logs Example: docker logs my-
container
Docker CLI for
Managing Containers
• docker exec
• Runs a command inside a running container
• Example: docker exec -it my-container bash
• docker inspect
• Provides detailed information about a container
• Example: docker inspect my-container
• docker cp
• Copies files/folders between the container and the host
• Example: docker cp my-container:/path/to/file /host/path
Docker CLI for
Managing Containers

• docker stats
• Displays real-time resource usage statistics of running containers
• Example: docker stats my-container
Images Stored
in the Docker
Registry
Containers
Isolated
from Each
Other
Working
with Docker
• Certainly!
• Here's an example slide on the topic "Working
with Docker": Slide X: Working with Docker
Building Docker
Images
• Building Docker images enables developers to package their applications
along with their dependencies, ensuring consistent and reproducible
environments for deployment
• Feel free to customize the slide content and design as per your
presentation style and requirements
Running Docker
Containers
• Let's consider a simple example where we want to containerize a web
application using Docker
• # Expose port 80 for web traffic # Start the nginx server This command
builds the Docker image and tags it with the name 'my-web-app'
• This command creates a container from the 'my-web-app' image, runs it in
detached mode , and maps port 8080 of the host machine to port 80 of
the container
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Amazon Web Services

Introduction to Amazon Sagemaker

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Learning Pathway of MLOps :
Building, training, and
deploying models from
SageMaker Studio

SageMaker infrastructure
fundamentals

Pre-built and custom

algorithms

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Sagemaker :
Data
Business
scientists
analysts
Make ML predictions Prepare data and
using a visual interface build, train, and
with Amazon deploy ML models
SageMaker Canvas with Amazon
Amazon SageMaker Studio

SageMaker
Infrastructure, tools,
visual interfaces,
workflows, orchestration,
and collaboration

ML
engineers
Deploy and manage
models at scale
with Amazon
SageMaker MLOps

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Adjusting to SageMaker workflows

Work from the

notebook, but not
on the notebook

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Notebook-based experiments: Familiar but limiting

Difficult to right-size for both light and heavy tasks

[1]: Data loading

[2]: Data exploration
How do we track and reproduce experiments?
[3]: Data cleaning
[4]: Data preparation
[*]: Model training
[ ]: Model evaluation How do I deploy and integrate my model?
[ ]: Model deployment?

Notebook instance / server / laptop

Need team environment management tools

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
What is AWS Sagemaker ?

Amazon SageMaker is a fully managed

machine learning service.
• Easily build and train machine
learning models
• deploy into a production-ready
hosted environment.
• Provides Jupyter authoring notebook
instance for easy access to your data
sources for exploration and analysis

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
References :

https://docs.aws.amazon.com/sagemaker/latest/dg/whatis-features-
alpha.html

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
IMMERSION DAY – GETTING STARTED WITH AMAZON SAGEMAKER

Amazon Web Services

Why Amazon Sagemaker ?

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
SageMaker separates notebook from job infrastructure
SageMaker containerized runtime

Processing jobs Training jobs Data typically on Amazon S3

Batch inference Online inference

API / Python SDK

[1]: Data exploration

[2]: Writing processing/training scripts
[*]: Running + comparing experiments
[ ]: Managing models, endpoints, etc

SageMaker Studio / Notebook instance

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
SageMaker separates notebook from job infrastructure
SageMaker containerized runtime
Optimize and distribute per-task compute
Processing jobs Training jobs

Batch inference Online inference

Automatically track job history, metrics,

API / Python SDK
metadata, logs, lineage

[1]: Data exploration

[2]: Writing processing/training scripts Deploy to real-time APIs or scalable
[*]: Running + comparing experiments batch jobs without boilerplate code
[ ]: Managing models, endpoints, etc

SageMaker Studio / Notebook instance

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Next Journey :

Get started with Amazon Sagemaker

Intended for Learners of AWS MLOps Course by

Manifold AI Learning
Setting Up Sagemaker

Sagemaker Studio Setup

Intended for Learners of AWS MLOps Course by

Manifold AI Learning