This document provides an overview of artificial intelligence, machine learning, cloud computing and their applications. It discusses key concepts like different machine learning algorithms including neural networks, deep learning and natural language processing. It also highlights various use cases of AI in different industries like automotive, banking, supply chain etc. and the impact of AI on businesses. Finally, it covers cloud computing delivery and deployment models as well as best practices for data management in the cloud.
2. Agenda
• Introduction of Data Science
• Various Roles in Data Science
• Artificial Intelligence
• Machine Learning
• Natural Language Processing
• Impact of AI on Business
• AI Use Cases
• Cloud Computing
• Delivery Models
• Deployment Models
• Data Management on Cloud
• Steps of Cloud Migration
• Cloud Migration Challenges and Strategies
• Success Stories
8. Types of Machine Learning
• Unsupervised - Unsupervised learning is a type of machine
learning algorithm used to draw inferences from datasets
consisting of input data without labelled responses
• Supervised - Supervised learning is inferring from labelled
training data
9. Which customers are
most likely to buy the
product
1. Problem
Identification
C
Collect data from Social
Media, product
reviews, purchase
trends, shopping cart
behaviour
2. Data
Collection
Whether the customer
has bought the product
in last 6 months
3. Data
Labelling
How to rank the result
in terms of strong lead
or weak lead
6. Model
Evaluation
Criteria
Randomly split data
into train and test
5. Split Dataset
Transform raw data into
features
4. Feature
Extraction
Train the model
7. Model
Training
Explore the results
8. Analyse
Model Output I
Identify the customer
who are likely to buy
the product
9. Model
Deployment
ML Process
Steps
11. Use Cases of ML
• Automotive – Driverless cars
• Banking – Opportunity to product new market, balance risk and detect fraud
• Government – Enhance security and threat detection
• Manufacturing – Product defects, Predictive Maintenance, Operational Efficiencies.
• Retail – Continuous monitoring of consumer behaviour, Micro segmentation, product
recommendation
• Education – Career selection, Online teaching
• E-commerce – Recommendation System, Last Mile Delivery, Customer engagement
12. Challenges with Machine
Learning
• Data Integration and Preparation
• Feature Extraction - Combining the original features
(e.g., variables) into a smaller set of more
representative features
• Feature Selection - Selecting only the most meaningful
original features to be used in the modeling algorithm
• Feature Engineering - Combining pre-existing features
in a data set with one another or with features from
external data sources to create new features that can
make models more accurate
13. Artificial Neural Networks
• It’s a supervised learning system
built of neurons or perceptrons
• Neurons, connected with each
other via interconnected layers,
process data to drive information
• A shallow neural network has only
three layers
• An Input Layer
• Hidden Layer
• Output Layer
14. Types of ANN
14
FeedForward
• Unidirectional information flow
• No feedback loop
• Fixed number of inputs and output
FeedBack
• Bidirectional information flow via feedback
loop
FeedForward Feedback
15. Application of AI using Neural Network
15
1. Automobile guidance system - In the case of self-driving cars, AI can help with being the brains of the cars doing things like
automatically detecting people and other cars around the vehicle, staying in the lane, switching lanes, and following the
GPS to get to the final destination
2. Target tracking - Using AI to Recognize Objects in Milliseconds. Intelligence, surveillance, and reconnaissance platforms
such as satellites, UAVs, and autonomous systems must process data in milliseconds to transmit decision-making
information to the warfighter in real-time
3. Computer vision - A field of artificial intelligence (AI) that enables computers and systems to derive meaningful
information from digital images, videos and other visual inputs — and take actions or make recommendations based on
that information.
4. Loan Advisor - An AI-powered mortgage advisor tracks market changes in real time to ensure that all the information
available is fully up to date. An automated advisor can provide with realistic estimates of maximum loan amount, the
expected down payment, potential closing costs, and other expenses
5. Credit Application Evaluators – An AI powered credit application evaluator system will do risk assessment of issuing the
loan, Data handling from a very wide range of touch points, such as income sources, purchase patterns and overall financial
behaviours of customers, and early warnings to reduce the impact of bad loans.
16. Deep Learning Neural Network (DNN)
• Deep learning is the extension of ANN with added hidden layer to
solve complex problems with high accuracy
16
Types of DNN
• Convolution Neural Networks - CNN
• Recurring Neural Networks - RNN
• Long Short-Term Memory - RNN
17. CNN – Full Architecture
17
Reference - https://medium.com/technologymadeeasy/the-best-explanation-of-convolutional-neural-networks-on-the-internet-fbb8b1ad5df8
18. Application of CNN
18
• Image Tagging –This has important role in visual search as well by comparing the input image with database search for other
photos that have comparable credentials.
• Recommender Engines – This is the business application based on object identification and image classification. When we visit
Amazon, we see a message “you might also like” with list of suggestions by recommendation system of Amazon.
• Facial Recognition – The application deals with recognition of complex images such as faces, animals, insects etc. The first step
in facial recognition is object recognition. Then extract features of the object such as shape of nose, skin tone, presence of
scars, hair, eyes distance etc.
• Medical Image Computing – CNN helps in detecting anomalies in X-ray and MRI images with better accuracy. The technique
helps doctor with information which help them in taking quick informed decisions. The algorithm is trained on Pubic Health
Records of Patients.
• Number Plate Recognition – The technique is applied in controlling the traffic as well as road disturbance by capturing and
reading the number plates of vehicle whose drivers are not obeying traffic rules. The same technique is applied on road
surveillance, toll plazas to capture vehicle movement and give important information to traffic police in case any crime
happens
19. Recurring Neural Network
19
• RNN is a deep learning algorithm used for
processing sequential data
• It has short term internal memory to
remember its input
• The information cycles through a loop
• It has two input (Present and Recent Past)
• It applies weights to both present and recent
past inputs
• The most famous application of RNN is Siri
from Apple and Google Translation service
21. Natural Language Processing
Branch of Artificial Intelligence that deals with
interaction between human and machine.
NLP Objective
Read
Decipher
Understand
Make Sense
Reply
22. Use Case Of
NLP – Around
us
Language translation applications such as
Google Translate
Grammatical accuracy of texts (Microsoft
Word and Grammarly)
Interactive Voice Response
Personal Assistant – OK Google, Siri,
Cortana and Alexa
23. Relation Between AI, ML, Deep Learning & NLP
Deep
Learning
Artificial Intelligence
Machine Learning
NLP
29. Promising Use
Cases of AI in
Logistics
1. Automated Warehouses
• Demand Prediction of particular product
• Modify orders and deliver in-demand items to local
warehouse
• Lower transportation cost
• Ocado Success Story – Supermarket in United
Kingdom. It has developed an automated warehouse.
The system is based on a robot called ‘hive-grid-
machine.’ This robot can execute 65,000 orders per
week. ‘Hive-grid-machines’ main task is to move, sort,
and lift items inside the warehouse. Ocado’s
automated warehouse dramatically cuts labor and the
time for orders to be executed.
30. Promising Use
Cases of AI in
Logistics
2. Autonomous Vehicles
• Save time & money
• Reduce chances of accidents
• Warehouse ground vehicles and drones
• Rolls-Royce Success Story– Rolls-Royce is working with
Intel to develop self-driving ships. It released the
Intelligence Awareness system that is able to classify
all the nearby objects under the water. It can also
monitor the engine condition and recommend the best
routes.
31. Promising Use
Cases of AI in
Logistics
3. Back Office
• Artificial Intelligence with Robotic Process Automation
(RPA) provides the workers with an opportunity to
increase their quality of work
• It lowers costs and improves the accuracy and
timeliness of data for logistics companies.
• UIPath Success Story– Developed a robot that is able
to conquer approximately 99% of back office tasks
40. PublicCloud
a) The infrastructure in this cloud model is owned by the entity that delivers the cloud services, not by the consumer
b) There is no substantial upfront fee. It works in the model of pay-per-use service
c) Infrastructure on cloud is run and maintain by cloud service provider
d) Cloud resources are available on demand basis
e) Customer has less control on infrastructure setup on cloud
PrivateCloud
a) Its direct opposite to the public cloud model
b) Infrastructure is not shared with any other user
c) Customer has the best control over services, IT operations, policies etc
d) Suitable for storing corporate information to which only authorized staff have access
e) User can deploy license software
41. HybridCloud
a) It gives best of both public and private cloud model
b) Customer can host the application on safe private cloud environment and take
advantage of public cloud’s saving.
c) The model provide both flexibility and control on cloud environment
CommunityCloud
a) It allows systems and services to be accessible by a group of organizations
b) Works on integrating the services of different clouds to address the specific
needs of a community, industry, or business
c) The infrastructure of the community could be shared between the organization
42. Data Management on Cloud
Easy accessibility of data on cloud must be balanced with protection to ensure maximum business value. We need to take
care of 5 important points to maintain that balance
─ Resting Data – When data rest in storage is considered as Resting Data. The data should be behind layers of security
along with proper encryption. Employees are the soft target for hackers. Encryption saves data not only from malicious
intent but also from careless action
─ Accessing Data – The access of the data must be under control environment. Any person requesting access to data
must be authenticated and every data transaction should be recorded so you can audit later if necessary.
─ Data in Transit - Create a secure, authenticated and encrypted tunnel between the authenticated user and device and
the data they’re requesting. For ex Virtual Private Network
─ Data Arrival – Setup mechanism to check data integrity and clear audit trail. This will protect the system from malware
and phishing attack
─ Backup and Recovery - Do not store your backups in the same cloud account where your production data resides.
Leverage multiple cloud accounts to segregate your backup data from your production data.
43. Steps of Cloud Migration
─ Outline Reasons for Moving to the Cloud – Outline business objective such as reduce costs, gain new features, leverage real-
time data, analytics, scalability
─ Determine Organizational Cloud-Readiness - Conduct a comprehensive business and technical analysis of current
environment, apps, and infrastructure.
─ Choose a Cloud Vendor - The right platform for your business will depend on specific requirements, the architecture of the
applications moving to the cloud, integrations etc
─ Create Cloud Roadmap - Outline which components will make the move first based on business priority and migration
difficulty.
─ Get Application Cloud Ready – Choose between Lift and Shift (Rehost) and Rearchitect (Refactor)
─ Migrate Data - Audit the data to prevent any unexpected issues, clean-up of any identified concerns, putting controls in place
to ensure data quality, and proper governance through tracking and monitoring.
─ Ongoing Upkeep – Flip the switch from testing to production. Migrate a set of test users over to the new environment before a
full launch to identify any issues that were missed through deployment and initial testing.
─
44. Cloud Migration Challenges
─ Lack of Strategy–The organization must have a clear business case for each workload it migrates to the cloud.
─ Cost Management– Cloud environments are dynamic and costs can change rapidly as new services are adopted and application
usage grows.
─ Vendor Lock-in–There is a high switching cost to migrate workloads from one cloud to another cloud service provider. A detail
due diligence is required before finalising the cloud hosting environment.
─ Data Security and Compliance– In the shared responsibility model cloud service provider take responsibility of securing the
cloud infrastructure and customer is responsible for securing data and workloads.
45. Cloud Migration Strategies – 5 R
─ Rehost– Also called “Lift and Shift” which means redeploy existing data and applications on the cloud server.
─ Refactor – Also called “Lift, tinker and Shift” which means tweak the applications as per the new cloud environment.
─ Rearchitect–This strategy is best suited for companies who decide to migrate on cloud at later stage. This allows company to
divide the application into several functional components that can be individually adapted and further developed.
─ Rebuild–Rebuild involves removing existing code and redesigning the application in the Cloud, after which you can utilize
innovative features on the Cloud provider’s platform.
─ Replace - A “Replace” migration strategy completely replaces an existing application with SaaS.
46. Best Practices of Migration to Cloud
─ Understand the scope and life cycle of the application
─ Review and choose an IT partner that can meet SLAs.
─ Properly manage software licensing
─ Monitor the Cloud usage periodically
─ Leverage service provider support
47. Success Stories – NSW Health
NWS Health Pathology Solution - It is one of Australia’s largest public health sector pathology providers. It
operates over 60 laboratories and conducts more than 61 million tests each year.
Business Problem – The lab does Covid-19 test of patients. The lab was taking 10 days to deliver the result to
the patients, which was elongating the isolation time and increase patient anxiety. The lab wanted to automate
the SMS notification service so that the results of Covid-19 test can be shared with patient with in 2 hours.
Solution – The company applied Amazon Redshift cloud data warehouse service and Amazon Connect
omnichannel cloud contact center to help users provide superior customer service at a lower cost because of its
pay-as-you-go pricing.
Benefits of Migrating on Cloud
─ Provides COVID-19 test results in less than 2 hours
─ Saved over 1 million clinical hours for frontline workers
─ Delivered over 4.25 million COVID-19 test results
─ Built solution and registered first patient in 2 weeks
─ Registered 87% of patients who were tested at state-run facilities in NSW
48. Success Stories – Paytm
Paytm - Paytm is the consumer brand of India’s leading mobile internet company, One97 Communications. The
brand is one of India’s largest financial services companies, offering full-stack payments and financial solutions to
consumers, offline merchants, and online platforms. Today, the company serves millions of merchants and
customers on its platform in India.
Business Problem – The company wanted to develop further capabilities in tech solution by adding facial
recognition and other machine learning capabilities on OCR problems, and gradually replace existing third-party
software with AWS solutions.
Solution – By using Amazon Textract, Paytm extracts user data from images of complex identity documents with
97 percent accuracy. Once the information is captured, Amazon Textract helps to identify image noise in real
time, allowing Paytm to immediately notify onsite agents to retake users’ identity document pictures when
necessary, saving both parties the inconvenience of repeat visits..
Benefits of Migrating on Cloud
─ Reduced time required for user KYC from days to minutes
─ Deployed KYC solution in one hour
─ Reduced costs by 75 percent
─ Better customer and merchant experience
49. Success Stories – ERGO Insurance
ERGO Insurance Singapore - a registered general insurer in Singapore and a wholly owned subsidiary of ERGO
Group AG, which in turn is a fully owned subsidiary of Munich Re Group. With a background in commercial
insurance, the company offers innovative new solutions for commercial and private insurance customers.
Business Problem – Monetary Authority of Singapore (MAS) mandated stricter cyber hygiene requirements for
financial institutions (including insurance providers) that ERGO had to follow. In order to comply the new
regulations, the company need to upgrade on-premise hardware which incur huge cost for the company
Solution – The company migrated on AWS cloud to achieve secure yet cost-efficient infrastructure without
making huge investment in hardware. The company chose Blazeclan as its AWS Partner to help with the
migration and ongoing operations in the AWS Cloud.
Benefits of Migrating on Cloud
─ Complies with the Monetary Authority of Singapore cyber hygiene and technology risk-management
requirements
─ Achieves at least 99% uptime and low website latency
─ Receives 24x7 support from AWS Partner
─ Saves 4 hours daily with automated database backups
─ Improves efficiency with seamless data transfers and integrations
50. Success Stories – Thompson Reuters
Thomson Reuters- It is a news and information services company that provides solutions for tax, law, media, and
government across 100 countries. Its business unit ONESOURCE GTM helps clients meet compliance for their
imports and exports through its software.
Business Problem – The had a time-consuming, manual DR process in place that required a full team of
engineers to manage two separate on-premises data centers, and it did not provide the data protection and
recovery times the company wanted.
Solution – The company adopt AWS Elastic Disaster Recovery (CloudEndure Disaster Recovery), which minimizes
downtime and data loss with fast, reliable recovery of on-premises and cloud-based applications using
affordable storage, minimal compute, and point-in-time recovery.
Benefits of Migrating on Cloud
─ Replicated over 120 TB of data from 300 servers
─ Set up a recovery site in the cloud
─ Eliminated its manual DR process
─ Optimized spending on its disaster recovery process
─ Enhanced its security and compliance
51. Major Trends in Cloud Computing
Cloud First Approach Will Reign -
Due the challenges companies have faced during Covid-19, we see a strong adoption of cloud-based setup
for both business critical and heavy workloads.
Cloud Native Application Will Thrive
Cloud is coming with plenty of pre build softwares that companies will no longer have incentive to build a
solution from scratch. For example AWS or Google Facial recognition services are way more advanced in
comparison to inhouse build solution.
Data analysis will grow on Cloud
In order to extract insights from data companies need a supporting infrastructure and platform that will
scale as need increases. This makes Data Analysis as a good service to be offered by Cloud Service
Providers.