AI-900

Chroma DB
Azure AI Fundamentals

- AlokKS
 AI-900

Exam Topics

• Describe Artificial Intelligence workloads

and considerations

• Describe fundamental principles of

machine learning on Azure

• Describe features of computer vision

workloads on Azure

• Describe features of Natural Language

Processing (NLP) workloads on Azure

• Describe features of generative AI

workloads on Azure

- AlokKS
 AI-900

What is AI

Artificial Intelligence (AI) refers to the

simulation of human intelligence in
machines that are programmed to think,
learn, and perform tasks that typically
require human-like cognitive processes.

- AlokKS
 AI-900

Machine Learning (ML)

A subset of AI where computers are trained

to learn from data and make predictions or
decisions without being explicitly
programmed for each specific task.

A machine learning model might be used to

predict customer churn based on historical
data.

- AlokKS
 AI-900

Deep Learning (DL)

A more advanced form of machine learning

based on artificial neural networks that
attempt to simulate the way the human
brain works. Deep learning excels at tasks
like image and speech recognition.

A deep learning system can identify objects

in images, such as distinguishing between
cats and dogs.

- AlokKS
 AI-900

Natural Language Processing

(NLP)
The branch of AI that focuses on enabling
machines to understand, interpret, and
generate human language.

Virtual assistants like Siri and Alexa use NLP

to process and respond to voice commands.

- AlokKS
 AI-900

Types of ML

Reinforcement Learning

Supervised Learning Unsupervised Learning

Self-supervised Learning Semi-supervised Learning

- AlokKS
 AI-900

Supervised Learning

The algorithm is trained on a labeled

dataset, where each data point is associated
with a correct output. The algorithm learns
to map inputs to outputs by identifying
patterns in the data.

Common Tasks - Regression,Classification

- AlokKS
 AI-900

Unsupervised Learning

The algorithm is trained on an unlabeled

dataset, where the data points have no
associated output labels. The algorithm
discovers hidden patterns and structures
within the data.

Common Tasks - Clustering,Dimensionality

reduction

- AlokKS
 AI-900

Reinforcement Learning

The algorithm learns by interacting with an

environment. It receives rewards or
penalties for its actions and learns to make
decisions that maximize the cumulative
reward.

Common Tasks - Robotics,Game playing

- AlokKS
 AI-900

Features

An individual measurable property or

characteristic of the data that is used to
make predictions or classifications. Features
are the input variables that are fed into a
machine learning model to help it learn
patterns and make decisions.

For a house price prediction model ,Size of

the house,Number of bedrooms.Location,
Age of the house,Lot size,Proximity to
schools etc.

- AlokKS
 AI-900

Labels

the output or target variable that the model

is trying to predict or classify. It is the value or
category associated with the input data
(features) that we use to "train" the model.

Labels: The actual price of the house

- AlokKS
 AI-900

Training & Testing Data

The training data is used to teach the model.

It includes both the features (input variables)
and labels (target or output) for supervised
learning tasks.

The testing data is used to evaluate how well

the trained model performs. It provides a
way to check the model's ability to
generalize to new, unseen data.

- AlokKS
 AI-900

Model Training

A machine learning algorithm uses a set of

input data(features) and their
corresponding known outcomes (labels) to
adjust its internal parameters, improving its
ability to predict the correct output when
given new, unseen data.

Model-A mathematical representation or

function that maps input data to predicted
outputs.

- AlokKS
 AI-900

Model Evaluation

Process of assessing the performance of a

machine learning model after it has been
trained. The goal is to determine how well
the model generalizes to new, unseen data,
and to understand its strengths and
weaknesses.

- AlokKS
 AI-900

Model Evaluation

• Accuracy:The proportion of correct

predictions made by the model out of all
predictions.
• Precision:The proportion of true positive
predictions (correctly predicted positive
cases) out of all predicted positive cases.
• Sensitivity:The proportion of true positive
predictions out of all actual positive cases.
• F1-Score: The harmonic mean of precision
and recall, which balances the two metrics.
• Confusion Matrix:A table that summarizes
the performance of a classification model
by comparing predicted labels with actual
labels.
• Mean Absolute Error:The average of the
absolute errors between the predicted and
true values.
- AlokKS
 AI-900

Confusion Matrix

• True Positives(TP): The model correctly

identified 80 spam emails as spam.
• False Negatives(FN): The model incorrectly
classified 20 spam emails as "not spam".
• False Positives(FP): The model incorrectly
classified 10 non-spam emails as spam.
• True Negatives(TN): The model correctly
identified 90 non-spam emails as
non-spam.

- AlokKS
 AI-900

Azure ML

A cloud-based service provided by Microsoft

Azure that enables data scientists,
developers, and organizations to build, train,
and deploy machine learning models at
scale. It provides a comprehensive suite of
tools, APIs, and services that streamline the
process of developing machine learning
models, facilitating collaboration,
automating workflows, and operationalizing
machine learning models efficiently.

- AlokKS
 AI-900

Resources

Setup or infrastructural resources needed to

run a machine learning workflow

Workspace

Datastore
Compute

- AlokKS
 AI-900

Azure ML Workspace

Provides a unified and collaborative

environment for managing and tracking
machine learning (ML) experiments, models,
datasets, and compute resources. It is a
Central place to manage all machine
learning resources.

- AlokKS
 AI-900

Compute

A compute is a designated compute

resource where you run your job or host your
endpoint.

Compute instance
Serverless compute
Compute cluster
Attached compute Inference cluster

- AlokKS
 AI-900

Datastore

To securely keep the connection information

to your data storage on Azure, so you don't
have to code it in your scripts.

Azure Blob Container

Azure Data Lake
Azure File Share

Azure Data Lake Gen2

- AlokKS
 AI-900

Assets

Created using Azure Machine Learning

commands or as part of a training/scoring
run. Assets are versioned and can be
registered in the Azure Machine Learning
workspace.

Model

Data

Component

Environment

- AlokKS
 AI-900

Environment

Set of software dependencies, packages,

and configurations required to run machine
learning experiments, training jobs, or
model deployments.

Curated: Pre-configured environments

provided by Azure ML, containing
collections of Python packages and settings
for popular machine learning frameworks.
These are ready to use "as is," offering quick
setup and faster deployment times.

Custom : These environments allow users to

define their own dependencies. You can
create custom environments using docker.

- AlokKS
 AI-900

Component

An Azure Machine Learning component is a

self-contained unit of code that performs a
specific task in a machine learning pipeline,
such as data processing, model training, or
scoring. Components are the building
blocks of advanced pipelines. Similar to a
function, a component has a name,
parameters, takes input, and produces
output. These components help modularize
and streamline the machine learning
workflow, enabling reusable and scalable
processes.

- AlokKS
 AI-900

Data

uri_folder and uri_file : a location in storage

that can be easily mapped to the filesystem
of a compute node in a job by either
mounting or downloading the storage to
the node.

mltable is an abstraction for tabular data

that is to be used for AutoML Jobs, Parallel
Jobs, and some advanced scenarios. If you're
just starting to use Azure Machine Learning
and aren't using AutoML, we strongly
encourage you to begin with URIs.

- AlokKS
 AI-900

Datasets

To manage and use data for training and

testing models.

Azure ML Workspaces help you manage

datasets by storing, sharing, and versioning
them across projects. You can register
datasets in the workspace to make them
accessible to different team members.

- AlokKS
 AI-900

Model

Consist of one or more binary files that

represent a machine learning model and
any corresponding metadata.

Azure Machine Learning supports three

types of storage format for models:

mlflow_model

custom_model triton_model

- AlokKS
 AI-900

Azure ML Pipelines

Enable you to create, manage, and

automate end-to-end machine learning
workflows.
Azure ML Pipelines allow you to create
automated workflows for ML tasks. A
pipeline consists of a series of steps (data
preparation, model training, evaluation) that
are executed in sequence or in parallel. You
can automate the entire process, from data
ingestion to model deployment.

- AlokKS
 AI-900

Automated Machine Learning

(AutoML)
Automatically building and tuning models.

• Automated Data Preprocessing

• Model Selection
• Hyperparameter Tuning
• Feature Engineering
• Multiple Algorithms Support

- AlokKS
 AI-900

Model Deployment

The process of making a trained machine

learning model available for use in
real-world applications, such as
providing real-time predictions,
batch predictions, or scoring
new data.

- AlokKS
 AI-900

Model Monitoring

• Model drift occurs when the model’s

performance degrades over time due to
changes in the input data distribution or
underlying patterns. Azure ML provides
tools to monitor data drift and concept
drift.
• Data Drift occurs when the statistical
properties of the input data change,
impacting model performance.
• Concept Drift refers to changes in the
relationship between input variables and
the target variable.

- AlokKS
 AI-900

Responsible AI

• Fairness
• Reliability and safety
• Privacy and security
• Inclusiveness
• Transparency
• Accountability

- AlokKS
 AI-900

Azure Cognitive Services

• Vision
◦ Computer Vision
◦ Face API
◦ Custom Vision
◦ Form Recognizer
• Speech
◦ Speech-to-Text (STT)
◦ Text-to-Speech (TTS)
◦ Speech Translation
• Language
◦ Text Analytics
◦ Language Understanding (LUIS)
◦ QnA Maker
• Decision
• Search

- AlokKS
 AI-900

Computer Vision

Computer vision is a field of artificial

intelligence that enables machines to
interpret and process visual information
from the world.

• Image Classification
• Object Detection
• Image Segmentation
• Facial Recognition
• Optical Character Recognition (OCR)

- AlokKS
 AI-900

Convolutional neural networks

(CNNs)
Deep learning models designed for
processing visual data like images and
videos. They consist of convolutional layers
that detect features (e.g., edges, textures),
pooling layers that reduce spatial
dimensions, and fully connected layers for
classification. CNNs automatically learn
features from raw data, making them ideal
for image classification, object detection,
and segmentation. They are
computationally efficient due to shared
weights in convolutional filters . CNNs are
widely used in applications like facial
recognition, medical imaging, and
autonomous driving, leveraging their ability
to recognize complex patterns in visual
data. - AlokKS
 AI-900

Transformers

A Transformer is a deep learning model

architecture primarily used for
sequence-based tasks, such as NLP. Unlike
traditional models like RNNs, Transformers
use self-attention mechanisms to process
all input tokens in parallel, capturing
relationships between words regardless of
distance. This attention mechanism assigns
weights to different parts of the input,
enabling the model to focus on relevant
context. Transformers are highly efficient
and scalable, excelling in tasks like machine
translation, text generation, and
summarization.

- AlokKS
 AI-900

Multi-modal models

Models designed to process and integrate

multiple types of data, such as text, images,
audio, and video, to improve performance
on tasks that require understanding across
different modalities. These models combine
features from diverse input sources,
enabling more robust and context-aware
predictions. For example, a multi-modal
model might use both text and images to
describe an object or understand a scene.
Popular examples include models like CLIP
(which connects images and text) and
GPT-4, which can process both textual and
visual information.

- AlokKS
 AI-900

Azure AI Vision

Azure service for building and deploying

computer vision applications. It enables
developers to analyze and interpret visual
content using pre-trained models or
custom models. Key services include the
Computer Vision API (for extracting
information from images, such as text,
objects, and scenes), Custom Vision (for
training personalized image classifiers), and
Face API (for facial recognition and emotion
detection). Azure AI Vision supports tasks
like object detection, OCR, image tagging,
and content moderation.

- AlokKS
 AI-900

Facial Recognition

• Face detection involves identifying

regions of an image that contain a human
face, typically by returning bounding box
coordinates that form a rectangle around
the face.
• Face analysis, facial features can be used
to train machine learning models to
return other information, such as facial
features such as nose, eyes, eyebrows, lips,
and others.
• Facial recognition is to identify known
individuals from their facial features.

- AlokKS
 AI-900

Azure AI Face service

- AlokKS
 AI-900

Optical Character Recognition

(OCR)
Converting different types of documents,
such as scanned papers, PDFs, or images,
into editable and searchable data (e.g.,
extracting text from scanned documents).

The service identifies key text elements

such as characters, words, and lines, and
can also detect and extract structured data
from forms.

• Create Azure AI Service resource

• Connect to Vision Studio

- AlokKS
 AI-900

Azure AI Language

• Text Analytics
• Language Understanding (LUIS)
• Translator
• QnA Maker
• Speech to Text
• Text to Speech

- AlokKS
 AI-900

Tokenization

The process of breaking down text into

smaller units, called tokens, which can be
words, subwords, or characters. In NLP,
tokenization helps convert raw text into a
structured form that a machine learning
model can process. For example, in the
sentence "I love AI," tokenization splits it
into the tokens ["I", "love", "AI"]. Tokenization
is essential for tasks like text analysis,
sentiment analysis, and language
modeling.

- AlokKS
 AI-900

Semantic language models

To understand and generate text based on

the meaning of words, phrases, and
sentences rather than just their syntactic
structure. These models focus on capturing
the relationships and context between
words to generate more coherent and
contextually relevant outputs.

- AlokKS
 AI-900

Entity recognition and linking

ERL is a key task in NLP that involves

identifying and classifying entities (e.g.,
names, dates, locations, organizations) in
text and linking them to corresponding
real-world references. Named Entity
Recognition (NER) extracts entities from
text, while Entity Linking connects these
entities to unique identifiers, often within a
knowledge base like Wikipedia or Wikidata.
For example, recognizing "Apple" as an
organization rather than a fruit.

- AlokKS
 AI-900

Sentiment analysis and opinion

mining
Techniques used to determine the
emotional tone behind a piece of text, such
as identifying whether a statement is
positive, negative, or neutral. Sentiment
analysis focuses on classifying the
sentiment expressed in a document, while
opinion mining dives deeper to extract
specific opinions or attitudes toward
entities (like products or services).

- AlokKS
 AI-900

Key phrase extraction

NLP service that identifies the most

relevant and important phrases in a text
document. It analyzes the content to
extract key terms, such as significant nouns
or noun phrases, that summarize the main
topics or themes of the document. This
service is useful for tasks like content
indexing, search optimization, and
summarization.

- AlokKS
 AI-900

Azure Language Studio

Portal for building, training, and deploying

NLP models. It simplifies the process of
creating and testing models for tasks such
as sentiment analysis, entity recognition,
key phrase extraction, and language
understanding. Language Studio offers an
intuitive interface, allowing users to
customize pre-built models or create new
ones with minimal code. It also provides
tools for data labeling, model evaluation,
and integration with other Azure services.

- AlokKS
 AI-900

Custom question answering

Enables you to define and publish a

knowledge base of questions and answers
with support for natural language querying.

• Create a Language resource

• Create a new project
• Edit the knowledge base
• Train and test the knowledge base
• Deploy your project

- AlokKS
 AI-900

Conversational language
understanding (CLU)
Component of NLP that enables machines
to understand and respond to human
language in conversational contexts. CLU
focuses on interpreting user inputs,
identifying intent and extracting relevant
entities . It is essential for building
intelligent conversational agents, such as
chatbots and virtual assistants. CLU models
are trained to handle complex, varied
language patterns, including slang, context,
and ambiguous expressions. Microsoft
Azure's Language Understanding (LUIS)
service is a popular tool for developing CLU
applications, allowing developers to build
custom models for specific conversational
scenarios.
- AlokKS
 AI-900

Language Understanding (LUIS)

Allows you to build custom language

models to understand user intents and
extract entities from user input.
• Create a Language resource
• Create a Conversational Language
Understanding App
• Create intents, utterances, and entities
• Train the model
• Deploy and test the model

- AlokKS
 AI-900

Azure AI Speech

Provides speech to text, text to speech, and

speech translation capabilities through
speech recognition and synthesis. You can
use prebuilt and custom Speech service
models for a variety of tasks, from
transcribing audio to text with high
accuracy, to identifying speakers in
conversations, creating custom voices, and
more.

• Speech recognition - the ability to detect

and interpret spoken input
• Speech synthesis - the ability to generate
spoken output

- AlokKS
 AI-900

Azure AI Translator

Key features include text translation,

language detection, document translation,
and real-time speech translation. The
service uses neural machine translation
(NMT) to provide accurate, fluent
translations, improving over time through
user feedback. Azure Translator also
includes customization options to tailor
translations to specific domains or
industries

- AlokKS
 AI-900

Azure AI Document Intelligence

A cloud-based form recognizer service that

leverages advanced machine learning to
extract text, key-value pairs, tables, and
structures from documents.

create either a Document Intelligence or

Azure AI services resource in your Azure
subscription.

- AlokKS
 AI-900

Azure AI Search

Search service that leverages AI to enhance

traditional search capabilities and enables
the creation of powerful, customizable
search experiences across websites, apps,
and enterprise data. Key features include
full-text search, filtering, and faceted
navigation, along with AI-powered
capabilities like semantic search for more
relevant results, automatic entity
recognition, and language understanding.

- AlokKS
 AI-900

Generative AI

A form of artificial intelligence in which

models are trained to generate new original
content based on natural language input.
It uses advanced models like Generative
Adversarial Networks (GANs) and
Transformers (e.g., GPT-3)

- AlokKS
 AI-900

Generative Adversarial Network

A class of machine learning frameworks

designed to generate new, synthetic data
samples that resemble a given dataset.

Generator: This network generates new

data samples. It tries to create data that is
indistinguishable from real data.

Discriminator: This network evaluates the

generated data. It tries to differentiate
between real and generated data.

- AlokKS
 AI-900

Embeddings

Embeddings are numerical representations

of text. They transform words, phrases, or
documents into vectors of numbers,
capturing the semantic meaning and
relationships between them.

Text to Vectors - Semantic Similarity-Cosine

- AlokKS
 AI-900

Attention

A technique used in deep learning models,

to improve the performance of models by
allowing them to focus on specific parts of
the input data when making predictions.

Input->Context ->Attension
Score->Weight-Output

- AlokKS
 AI-900

Copilots

An AI-powered assistant integrated into

Microsoft 365 apps like Word, Excel,
PowerPoint, and Outlook. It enhances
productivity and creativity by helping users
with tasks such as drafting and editing
documents, analyzing data, creating
presentations, and managing emails.

To customize Microsoft copilot or develop

custom copilots, Microsoft provides two
tools that you can use, Copilot Studio and
Azure AI Studio.

- AlokKS
 AI-900

Azure AI Studio

A web portal that brings together multiple

Azure AI-related services into a single,
unified development environment.

AI HUB

AI Studio

• Deploy LLM
• Test
• Add own data
• Evalualte Models
• Define filters

- AlokKS
 AI-900

Responsible GenAI

• Identify potential harms that are relevant

to your planned solution.

• Measure the presence of these harms in

the outputs generated by your solution.

• Mitigate the harms at multiple layers in

your solution to minimize their presence
and impact, and ensure transparent
communication about potential risks to
users.

• Operate the solution responsibly by

defining and following a deployment and
operational readiness plan.

- AlokKS
 Important Resources
• https://learn.microsoft.com/en-us/training/courses/ai-900t00
• https://medium.com/
• https://www.geeksforgeeks.org/
• https://www.wikipedia.org/
• and others

- AlokKS
Thank
Y U
- AlokKS