Build and Modernize Intelligent Apps
Lorenzo Barbieri
Daniela Tomoiaga
Generative AI makes apps truly intelligent
Modern apps
Modern app development principles
Microservices-based architecture
Rapid innovation with CI/CD
Paradigm
shift
Intelligent apps
Natural language interaction
Data-driven, personalized experiences
that improve over time
Quickly deliver new features
Progressing from now to next
Traditional app architectures and platforms
are harder/unfeasible to adapt to these new
opportunities
Traditional databases struggle with
geographicallydistributed unstructured data
and ensuring low latency access
App developers working with AI, data and
other product teams need a unified
experience across environments that
provides an easy integration between
services and ensures smooth handoffs
CIOs recognize that the foundations…
to support AI development must evolve*
Transformation
Cloud Apps + data AI
*2023 Survey conducted by MIT Technology Review Insights, of 600 CIOs and other technology leaders
Innovate on a pressure tested AI platform
AI-based
search with
ChatGPT
Biometric identity
verification
Personalized
recommendations
Copilot for Work
across Office 365
Copilot
assisted coding
Microsoft modernized our own flagship apps, and augmented with AI copilots for highly differentiated experiences
Azure powers OpenAI and ChatGPT
ChatGPT
Runs on Azure Kubernetes Service
Backed by Azure Cosmos DB
Developed on GitHub
Fastest
growing consumer
product in history
Build new Intelligent apps:
Develop AI powered applications optimized for cloud
Realize the power of intelligent apps for your business
1 Build new intelligent apps
2 Add AI capabilities to existing apps
3 Build your own copilot
4 Develop new cloud-native applications
5 Modernize existing applications
A common platform with the technology you need
Application services
Azure
Kubernetes
Service
Azure
App Service
Azure
Container
Apps
Azure
Functions
AI portfolio
Azure AI Services
(including Azure
OpenAI Service)
Azure
AI Studio
Managed databases
Azure
SQL family
Azure
PostgreSQL
Azure
Cosmos DB
Services to bootstrap AI and integrate
with best of cloud
Turing
Rich language understanding
Z-Code
100 languages translation
Florence
Breakthrough visual recognition
GPT-3/GPT-4
Human-like language generation
DALL-E
Realistic image generation
Codex
Advanced code generation
ChatGPT
Conversation generation
Azure AI
services
Vision
Speech
Language
Decision
OpenAI Service
Cognitive Search
Form Recognizer
Immersive Reader
Bot Service
Video Analyzer
Breakthrough research, to
AI tools for developers and
data scientists to augment
the work that people do,
freeing up time for more
creative tasks and
innovative thinking
Empower developers to combine the power of AI,
cloud-scale data, and cloud-native app development
Identity
RBAC
Monitoring
Policy
AKS Cluster
Service Bus Queue Cosmos DB
Azure OpenAI
or other Azure
AI services
order-service
store-admin
makeline-service
store-front
product-service
ai-service
Add AI capabilities to existing apps:
Modernize and enhance existing applications with AI
Adopt a better software development paradigm
Traditional
Monolithic apps
Waterfall/sequential
App server on
single machine
Many tools,
ad hoc automation
Designated deploys
Cloud-Native
Microservice/composite apps
DevOps and CI/CD, iterative releases
Containers, infrastructure abstraction, serverless
Managed databases, AI models, APIs
Always shipping code and business value
End-to-end application platform for intelligent apps
Community Security
Azure
Active
Directory
Azure
Policy
Microsoft
Defender
AI
Machine Learning
Azure AI Services Azure AI Studio
Development tools
GitHub
Azure API Management
Azure Container Registry
Visual Studio Code
…and more…
Azure
Kubernetes
Service
Azure
Container
Apps
Application platforms
Azure
Functions
Azure
App
Service
Managed databases
Azure
PostgreSQL
Azure
Cosmos DB
Azure
SQL family
Azure
MySQL &
MariaDB
Storage
Azure
Storage
Observability
Azure
Monitor
Azure App
Insights
Azure Managed
Grafana
Azure Arc
Cloud operations anywhere
Demos and reference apps
BYO copilot - solution architecture
With Reference
Application you will be
able to connect your
enterprise data sources to
build a ChatGPT like
conversational interface.
It can be used to search,
explore internal data,
power a new generation
of Support, Sales, Bots. By
using Azure Cosmos DB
you guarantee full control
over your data, extreme
availability, real time
global access and
immediate scalability.
Legend: Source Data Semantic Kernel
Orchestrated Call
Other Service Call
Source Data
+
E mbeddings,
Prompts & Completions
DATA
SOURCES
Batch / CDC / Stream
Azure Cosmos DB
User Interaction
CONSUMERS
Data Engineers
DBAs
End Users
A zure
Ope nAI
(E mbeddings
AP I)
Q&A Web
App
Files
Databases
Apps
Ask Question
1
Submit
Question &
History
2
1 Q&A interaction
1 Source data loading
1
Change Feed:
Source
Documents
2
Index Documents &
Embedding Vectors
4
Persist Q&A
6
Persist Embedding
Vectors
5
5 Request Completion
(Completions API)
Request Embeddings
(Embeddings API)
3
C ognitive
Se arch
(H ybrid Index)
Azure
Ope nAI
(Completions
AP I)
Request Question
Embedding
(Embeddings API)
3
Search for
context data
4
Load Q&A Session
0
Repo: https://github.com/Azure/Build-Modern-AI-Apps
Azure
Kubernetes
Azure
Kubernetes
Reference App: Medical Claims Processing & Handling
Repo link: https://github.com/Azure/Medical-Claims-Transaction-Processing-at-scale
Uses: Azure Cosmos DB, Azure Kubernetes Services (AKS), Azure OpenAI Services, Azure Synapse Analytics,
Event Hubs, Static Web App
Reference App: Real-time Transactions & Payments
Repo link: https://github.com/Azure/Real-time-Payment-Transaction-Processing-at-Scale
Uses: Azure Cosmos DB, Azure Kubernetes Services (AKS), Azure OpenAI Services, Static Web App
Semantic Kernel & Plug-Ins
Introducing Semantic
Kernel
• Open-source SDK to orchestrateAI
workflows in your applications
• Support for Open AI, Azure Open AI
and Hugging Face
• Available in C#, Python and Java
• Plugins
• Automatic orchestration with planner
• Memory and embeddings
Extend your application with plugins
• Plugins give you the option to:
• Enable your AI solution to integrate 3rd party platforms
• Easily reuse prompts and features across multiple applications
Semantic Functions
(prompt)
Native Functions OpenAI Plugins
Introducing the planner
https://github.com/qmatteoq/SemanticKernel-Demos
GPT-4 Multimodal Capabilities
The Dawnof LMMs: Preliminary
Explorations with GPT-4V(ision)
GRAZIE!
LinkedIn.com/in/geniodelmale
Connect with me on LinkedIn
LinkedIn.com/in/daniela-tomoiaga
Connect with me on
LinkedIn

Build and Modernize Intelligent Apps​

  • 1.
    Build and ModernizeIntelligent Apps Lorenzo Barbieri Daniela Tomoiaga
  • 2.
    Generative AI makesapps truly intelligent Modern apps Modern app development principles Microservices-based architecture Rapid innovation with CI/CD Paradigm shift Intelligent apps Natural language interaction Data-driven, personalized experiences that improve over time Quickly deliver new features
  • 3.
    Progressing from nowto next Traditional app architectures and platforms are harder/unfeasible to adapt to these new opportunities Traditional databases struggle with geographicallydistributed unstructured data and ensuring low latency access App developers working with AI, data and other product teams need a unified experience across environments that provides an easy integration between services and ensures smooth handoffs CIOs recognize that the foundations… to support AI development must evolve* Transformation Cloud Apps + data AI *2023 Survey conducted by MIT Technology Review Insights, of 600 CIOs and other technology leaders
  • 4.
    Innovate on apressure tested AI platform AI-based search with ChatGPT Biometric identity verification Personalized recommendations Copilot for Work across Office 365 Copilot assisted coding Microsoft modernized our own flagship apps, and augmented with AI copilots for highly differentiated experiences
  • 5.
    Azure powers OpenAIand ChatGPT ChatGPT Runs on Azure Kubernetes Service Backed by Azure Cosmos DB Developed on GitHub Fastest growing consumer product in history
  • 6.
    Build new Intelligentapps: Develop AI powered applications optimized for cloud
  • 7.
    Realize the powerof intelligent apps for your business 1 Build new intelligent apps 2 Add AI capabilities to existing apps 3 Build your own copilot 4 Develop new cloud-native applications 5 Modernize existing applications
  • 8.
    A common platformwith the technology you need Application services Azure Kubernetes Service Azure App Service Azure Container Apps Azure Functions AI portfolio Azure AI Services (including Azure OpenAI Service) Azure AI Studio Managed databases Azure SQL family Azure PostgreSQL Azure Cosmos DB
  • 9.
    Services to bootstrapAI and integrate with best of cloud Turing Rich language understanding Z-Code 100 languages translation Florence Breakthrough visual recognition GPT-3/GPT-4 Human-like language generation DALL-E Realistic image generation Codex Advanced code generation ChatGPT Conversation generation Azure AI services Vision Speech Language Decision OpenAI Service Cognitive Search Form Recognizer Immersive Reader Bot Service Video Analyzer Breakthrough research, to AI tools for developers and data scientists to augment the work that people do, freeing up time for more creative tasks and innovative thinking
  • 10.
    Empower developers tocombine the power of AI, cloud-scale data, and cloud-native app development Identity RBAC Monitoring Policy AKS Cluster Service Bus Queue Cosmos DB Azure OpenAI or other Azure AI services order-service store-admin makeline-service store-front product-service ai-service
  • 11.
    Add AI capabilitiesto existing apps: Modernize and enhance existing applications with AI
  • 12.
    Adopt a bettersoftware development paradigm Traditional Monolithic apps Waterfall/sequential App server on single machine Many tools, ad hoc automation Designated deploys Cloud-Native Microservice/composite apps DevOps and CI/CD, iterative releases Containers, infrastructure abstraction, serverless Managed databases, AI models, APIs Always shipping code and business value
  • 13.
    End-to-end application platformfor intelligent apps Community Security Azure Active Directory Azure Policy Microsoft Defender AI Machine Learning Azure AI Services Azure AI Studio Development tools GitHub Azure API Management Azure Container Registry Visual Studio Code …and more… Azure Kubernetes Service Azure Container Apps Application platforms Azure Functions Azure App Service Managed databases Azure PostgreSQL Azure Cosmos DB Azure SQL family Azure MySQL & MariaDB Storage Azure Storage Observability Azure Monitor Azure App Insights Azure Managed Grafana Azure Arc Cloud operations anywhere
  • 14.
  • 15.
    BYO copilot -solution architecture With Reference Application you will be able to connect your enterprise data sources to build a ChatGPT like conversational interface. It can be used to search, explore internal data, power a new generation of Support, Sales, Bots. By using Azure Cosmos DB you guarantee full control over your data, extreme availability, real time global access and immediate scalability. Legend: Source Data Semantic Kernel Orchestrated Call Other Service Call Source Data + E mbeddings, Prompts & Completions DATA SOURCES Batch / CDC / Stream Azure Cosmos DB User Interaction CONSUMERS Data Engineers DBAs End Users A zure Ope nAI (E mbeddings AP I) Q&A Web App Files Databases Apps Ask Question 1 Submit Question & History 2 1 Q&A interaction 1 Source data loading 1 Change Feed: Source Documents 2 Index Documents & Embedding Vectors 4 Persist Q&A 6 Persist Embedding Vectors 5 5 Request Completion (Completions API) Request Embeddings (Embeddings API) 3 C ognitive Se arch (H ybrid Index) Azure Ope nAI (Completions AP I) Request Question Embedding (Embeddings API) 3 Search for context data 4 Load Q&A Session 0 Repo: https://github.com/Azure/Build-Modern-AI-Apps Azure Kubernetes Azure Kubernetes
  • 16.
    Reference App: MedicalClaims Processing & Handling Repo link: https://github.com/Azure/Medical-Claims-Transaction-Processing-at-scale Uses: Azure Cosmos DB, Azure Kubernetes Services (AKS), Azure OpenAI Services, Azure Synapse Analytics, Event Hubs, Static Web App
  • 17.
    Reference App: Real-timeTransactions & Payments Repo link: https://github.com/Azure/Real-time-Payment-Transaction-Processing-at-Scale Uses: Azure Cosmos DB, Azure Kubernetes Services (AKS), Azure OpenAI Services, Static Web App
  • 19.
  • 20.
    Introducing Semantic Kernel • Open-sourceSDK to orchestrateAI workflows in your applications • Support for Open AI, Azure Open AI and Hugging Face • Available in C#, Python and Java • Plugins • Automatic orchestration with planner • Memory and embeddings
  • 21.
    Extend your applicationwith plugins • Plugins give you the option to: • Enable your AI solution to integrate 3rd party platforms • Easily reuse prompts and features across multiple applications Semantic Functions (prompt) Native Functions OpenAI Plugins
  • 22.
  • 23.
  • 24.
  • 28.
    The Dawnof LMMs:Preliminary Explorations with GPT-4V(ision)
  • 33.
    GRAZIE! LinkedIn.com/in/geniodelmale Connect with meon LinkedIn LinkedIn.com/in/daniela-tomoiaga Connect with me on LinkedIn