Nishant Thacker
Microsoft Confidential Decision support Decision Global business value derived automation from AI in 2022 will reach $3.9T Smart $products 3.9T Virtual agents
“Forecast: The Business Value of Artificial Intelligence, Worldwide, 2017-2025”, Gartner, April 2018. Where do you see your company today?
Where do you see your top competitors today?
Where do you see your company in the future?
Analytics Capabilities BASIC ADVANCED Marketing Sales Service Finance Operations Workforce
ProductProduct RecommendationRecommendation Lead Scoring Intelligent chatbots Financial Forecasting PredictivePredictive Maintenance Employee Insights
Customer Insights Sales Insights Virtual Assistants Cash Flow Forecasting DemandDemand Forecasting HR Insights
Churn Analytics Dynamic Pricing Waiting line optimization Risk Management Quality Assurance Resource Planning Product recommendation Predictive maintenance Demand forecasting
The average size of a single cart has Unplanned downtime results in cost Solar energy production is inconsistent decreased overruns
Provide personalized digital content to Predict when maintenance should be Align energy supply with the optimal shoppers performed markets
Increase cart size Minimize downtime Maximize revenue
ASOS delivers 15.4 million personalized Hybrid solution predicts onboard water Distributed power generation increases experiences with 33 orders per second usage, saving $200k/ship/year revenue by over €100 million Serving business users and end users with intelligent and dynamic applications
Build a unified and usable Train ML and DL models to Operationalize models and data pipeline derive insights distribute insights at scale Complexity of solutions Data silos Difficult to scale effectively
AndMany most options aren’t satisfiedin the marketplace with their current solutions Incongruent data types Performance constraints
Are satisfied with ease of Are satisfied with access to semi- Are satisfied with ability to scale to 46% use of analytics software 21% structured and unstructured data 28% handle unexpected requirements
“What it Takes to be Data-Driven: Technologies and Best Practices for Becoming a Smarter Organization”, TDWI, 2017.
Data AI Data AI
Data Modernization on-premises AI apps & agents
Data modernization to Azure Knowledge mining
Globally distributed data
Cloud Scale Analytics
Machine Learning on Azure
Domain specific pretrained models … To reduce time to market Vision Speech Language Search
Familiar Data Science tools To simplify model development PyCharm Jupyter Visual Studio Code Command line
Popular frameworks To build advanced deep learning solutions Pytorch TensorFlow Scikit-Learn Onnx
Productive services
To empower data science and development teams Azure Azure Machine Machine Databricks Learning Learning VMs
Powerful infrastructure To accelerate deep learning CPU GPU FPGA
From the Intelligent Cloud to the Intelligent Edge Infuse apps with powerful, pre-trained AI models
Customize easily and tailor to your needs
Language Vision
Computer Vision | Video Indexer | Face | Content Moderator Text Analytics | Spell Check | Language Understanding | Text Translation | QnA Maker
…
Speech Bing Search
Speech to Text | Text to Speech | Speech Translation | Speaker Recognition Big Web Search | Video Search | Image Search | Visual Search | Entity Search | News Search | Autosuggest Empower data science and development teams
Integrated data science & data engineering teams Individual data scientists Desktop solutions not adequate Desktop solutions adequate Need a unified big data & machine learning solution Need cloud for sporadic compute needs
+
Azure Databricks Azure Machine Learning Machine Learning VMs service Fast, easy, and collaborative Apache Spark™-based analytics platform
Built with your needs in mind Increase productivity
Optimized Apache Spark environmnet
Collaborative workspace
Build on a secure, trusted cloud Integration with Azure data services
Autoscale and autoterminate
Optimized for distributed processing
Scale without limits Support for multiple languages and libraries
Seamlessly integrated with the Azure Portfolio Bring AI to everyone with an end-to-end, scalable, trusted platform
Built with your needs in mind Boost your data science productivity
Automated machine learning
Managed compute
Increase your rate of experimentation DevOps for machine learning
Simple deployment
Tool agnostic Python SDK
Deploy and manage your models everywhere Support for open source frameworks
Seamlessly integrated with the Azure Portfolio Accelerate deep learning
CPUs GPUs FPGAs
General purpose machine Deep learning Specialized hardware learning accelerated deep learning D, F, L, M, H Series N Series Project Brainwave
Optimized for flexibility Optimized for performance From the Intelligent Cloud to the Intelligent Edge
Train and deploy Train and deploy
Track models in production Capture model telemetry Retrain models
Deploy Quickly and easily build, train, deploy and manage your models
…
Domain Specific Pretrained Models Vision Speech Language Search To reduce time to market
Familiar Data Science tools To simplify model development PyCharm Jupyter Visual Studio Code Command line
Popular Frameworks To build machine learning and deep learning solutions ONNX Pytorch TensorFlow Scikit-Learn
Productive Services
To empower data science and development teams Azure Azure Machine Machine Databricks Learning Learning VMs
Powerful Infrastructure To accelerate deep learning CPU GPU FPGA
From the Intelligent Cloud to the Intelligent Edge
Ingest Store Prep and train Serve Model Serving
Azure Kubernetes Azure Event service Hubs Streaming data Operational Apps Databases
Cosmos DB, SQL DB
Azure Databricks Azure Machine Learning service
Ad-hoc Analysis Power BI
Batch data Azure analysis Azure Data Factory Azure SQL data Azure Data Lake Storage warehouse services PREP & TRAIN
A
B
C
Collect and prepare data Train and evaluate model Operationalize and manage
Azure Data Factory Azure ML Services Azure Databricks Azure Databricks Azure Databricks Azure Data Azure Blob Azure Factory Storage Databricks
Ingest Store Understand and transform
Connect to data Leverage best-in-class Scale from any source analytics capabilities without limits
• Integrate with all of your data sources • Leverage open source technologies • Build in the language of your choice • Create hybrid pipelines • Collaborate within teams • Leverage scale out topology • Orchestrate in a code-free environment • Use ML on batch streams • Scale compute and storage separately Tools Automated ML Azure Azure ML Databricks service Azure ML service
Machine learning Automated ML capabilities
Infrastructure Azure Azure ML Databricks service
Scale out clusters
Simplify model Scale compute resources Quickly determine the development to meet your needs right model for your data
• Collaborate in interactive workspaces • Easily scale up or scale out • Determine the best algorithm • Access a library of battle-tested models • Autoscale on serverless infrastructure • Tune hyperparameters to optimize models • Automate job execution • Leverage commodity hardware • Rapidly prototype in agile environments Azure Azure ML service ML service AKS ACI
Containers Docker
Model MGMT, experimentation, Train and evaluate models and run history IoT edge
Bring models Proactively manage Deploy models to life quickly model performance closer to your data
• Build and deploy models in minutes • Identify and promote your best models • Deploy models anywhere • Iterate quickly on serverless infrastructure • Capture model telemetry • Scale out to containers • Easily change environments • Retrain models with APIs • Infuse intelligence into the IoT edge Data Science VMs PyTorch
Keras
MS Cognitive Toolkit Azure ML AI Ready Virtual Machines service SDK
TensorFlow
Notebooks, IDEs, Command Line AML Compute
Scale out clusters
Streamline Scale compute Quickly evaluate AI development efforts resources in any environment and identify the right model
• Leverage popular deep learning toolkits • Choose VMs for your modeling needs • Run experiments in parallel • Develop your language of choice • Process video using GPU-based VMs • Provision resources automatically Ready-to-use clusters with Azure Databricks Runtime for ML
Tools Frameworks Infrastructure Use HorovodEstimator via a native Full Python and Scala support for Leverage powerful GPU-enabled VMs runtime to enable build deep learning transfer learning on images pre-configured for deep neural models with a few lines of code network training
Load images natively in Spark Seamlessly use TensorFlow, Microsoft DataFrames to automatically decode Cognitive Toolkit, Caffe2, Keras, and Automatically store metadata in Azure them for manipulation at scale with more Database with geo-replication for distributed DNN training on Spark fault tolerance
Simultaneously collaborate within Use built-in hyperparameter tuning notebooks environments to via Spark MLLib to quickly optimize Improve performance 10x-100x over streamline model development the model traditional Spark deployments with an optimized environment
Azure Databricks Remote support for other IDEs outside of native notebooks MLFlow for better DevOps with Azure Databricks and other ML pipelines Azure Machine Learning Python SDK support for popular IDEs & notebooks, including Azure Databricks Azure Machine Learning managed compute capabilities Introduce new models for FPGA scoring
Robust ONNX support - runtime engine in AML, model operationalization in SQL Server Automated machine learning Deploy and manage models to IoT edge Extend Machine Learning services to SQL DB AZURE ML SERVICE AZURE DATABRICKS
Bring AI to the edge Accelerate processing with the fastest Apache Spark engine
Increase your rate of experimentation Integrate natively with Azure services
Deploy and manage your models everywhere Access enterprise-grade Azure security
Leverage your favorite deep learning frameworks
TensorFlow MS Cognitive Toolkit PyTorch Scikit-Learn ONNX Caffe2 MXNet Chainer Get started quickly with Developer AI
© Microsoft Corporation Azure bot service + cognitive services
Start Bing search Text State 1 State 2 Dynamics365 Speech Bot framework State 3 SDK dialog manager On-premises Speech IFTTT
QnA Maker Dispatch
Image Language Translator understanding
© Microsoft Corporation Cognitive services Azure search Bot services
Use pre-built AI services Get up and Speed development with a purpose-built to solve business problems running quickly environment for bot creation
Map complex Reduce complexity with Infuse intelligence into your bot using information and data a fully-managed service cognitive services
Allow your apps to Use artificial intelligence Integrate across multiple process natural language to extract insights channels to reach more customers
Create a seamless developer experience across desktop, cloud, or at the edge using Visual Studio AI Tools © Microsoft Corporation Conversational agents Intelligent apps Business processes
Transform your engagements with Leverage AI to create the future Transform critical business customers and employees of business applications processes with AI
Of customer interactions 75% Applications to include Of enterprises using 95% powered by AI bots by 2025 AI by the end of this year 85% AI by 2020
© Microsoft Corporation Drones Smart factory Voice-activated speakers
Smart kiosks Smart kitchen Smart cameras
Smart bath
© Microsoft Corporation Empowering people with tools that amplify human capability
Employment Modern life Human connection Facilitating development Delivering personalized experiences Providing equal access to of professional skills to improve independence information and opportunity
How Microsoft How Helpicto How RIT is helping the is helping is helping the visually-impaired non-verbal children hearing-impaired
© Microsoft Corporation
Effective customer Decision services Risk and revenue Risk and compliance Recommendation engagement management management management engine
Customer profiles Customer segmentation Transaction data CRM Clickstream data Credit history CRM data Demographics Credit Products Transactional data Credit data Purchasing history Risk Services LTV Market data Trends Merchant records Customer service data Loyalty Products and services
Customer Financial Risk, fraud, Credit Marketing analytics modeling threat detection analytics analytics
Customer 360 Commercial/retail banking, Real-time anomaly detection Enterprise DataHub Recommendation engine degree evaluation securities, trading and investment models Card monitoring and fraud Regulatory and Predictive analytics and Customer segmentation detection compliance analysis targeted advertising Decision science, simulations Reduced customer churn and forecasting Security threat identification Credit risk management Fast marketing and multi- channel engagement Underwriting, servicing and Investment recommendations Risk aggregation Automated credit analytics delinquency handling Customer sentiment analysis Insights for new products
Faster innovation Improved consumer Enhanced customer Transform growth Improved customer for a better outcomes and experience with with predictive engagement with customer experience increased revenue machine learning analytics machine learning DNA Real world Image Sensor Social data sequences analytics deep learning data listening
FAST-Q HL7/CCD MRI Readings Social media BAM 837 X-RAY Time series Adverse events SAM Pharmacy CT Event data Unstructured VCF Registry Ultrasound Expression EMR
Genomics and Clinical and GPU image IoT device Social precision medicine claims data processing analytics analytics
Single cell sequencing Claims data warehouse Graphic intensive workloads Aggregation of Real-time patient feedback streaming events via topic modelling Biomarker, genetic, variant Readmission predictions Deep learning using and population analytics Tensor Flow Predictive maintenance Analytics across Efficacy and comparative publication data ADAM and HAIL on Databricks analytics Pattern recognition Anomaly detection Prescription adherence Market access analysis
Faster innovation Improved outcomes Diagnostics Predictive analytics Improved patient for drug and increased leveraging transforms quality communications development revenue machine learning of care and feedback Personalized Effective Information Inventory Consumer engagement recommendations customer retention optimization allocation analysis
Customer profiles Customer profiles Consumption logs Transactions Content metadata Viewing history Online activity Clickstream and devices Subscriptions Ratings Online activity Content distribution Marketing campaign Demographics Comments responses Content sources Services data Credit data Social media activity Channels
Content Customer churn Recommendation Predictive Sentiment personalization prevention engine analytics analysis
Personalized viewing and Quality of service and Ad effectiveness Predict audience interests Demand-elasticity engagement experience operational efficiency Content monetization Network performance Social network analysis Click-path optimization Market basket analysis and optimization Fraud detection Promotion events Next best content analysis Customer behavior analysis Pricing predictions time-series analysis Information-as-a-service Improved real time Click-through analysis Nielsen ratings and Multi-channel marketing ad targeting High value user engagement projections attribution Mobile spatial analytics
Faster innovation Improved consumer Enhance user Predictive Improved consumer for customer outcomes and experience with analytics engagement with experience increased revenue machine learning transforms growth machine learning Recommendation Effective customer Inventory Inventory Consumer engine engagement optimization allocation engagement
Customer profiles Shopping history Demand plans Demographics Historical sales data Shopping history Online activity Forecasts Buyer perception Price scheduling Online activity Floor plans Sales history Consumer research Segment level price changes Social network analysis App data Trends Market/competitive analysis Local events/weather patterns
Next best and Store design and Data-driven stock, Assortment Real-time pricing personalized offers ergonomics inventory, ordering optimization optimization
Customer 360/consumer Path to purchase Predict inventory positions Economic modelling Demand-elasticity personalization and distribution In-store experience Optimization for foot traffic, Personal pricing schemes Right product, promotion, Fraud detection Online interactions at right time Workforce and manpower Promotion events optimization Market basket analysis Flat and declining categories Multi-channel promotion Multi-channel engagement
Omni-channel Faster innovation Improved consumer shopping Predictive Improved consumer for customer outcomes and experience with analytics engagement with experience increased revenue machine learning transforms growth machine learning Effective customer Recommendation Risk and fraud Campaign reporting Brand promotion and engagement engine analysis analytics customer experience
Customer profiles Customer segmentation Transaction data CRM Social media Online history CRM data Demographics Merchant records Online history Transaction data Credit data Purchasing history Products Customer service data Loyalty Market data Trends Services Marketing data
Customer value Next best and Risk and fraud Sales and campaign Sentiment analytics personalized offers management optimization analysis Customer 360, segmentation Right product, promotion, Real-time anomaly detection Optimizing return on ad Opinion mining/social aggregation and attribution at right time spend and ad placement media analysis Fraud prevention Audience modelling/index Real time ad bidding platform Multi-channel promotion Deeper customer insights report Customer spend and Personalized ad targeting risk analysis Ideal customer traits Customer loyalty programs Reduce customer churn Ad performance reporting Data relationship maps Optimized ad placement Shopping cart analysis Insights for new products Historical bid opportunity as a service
Faster innovation Improved outcomes Risk management Predictive Improved customer for customer and increased with machine analytics engagement with growth revenue learning transforms growth machine learning Upstream optimization, Grid operations, asset Supply-chain Risk Recommendations maximize well life inventory optimization optimization optimization engine
Field data Sensor stream data Transaction data Sensor stream data Clickstream data Asset data UAVs images Demographics Transport Products Demographics Inventory data Purchasing history Retail data Services Production data Production data Trends Grid production data Market data Refinery tuning parameters Competitive data Demographics
Digital oil field/ Industrial Supply-chain Safety and Sales and marketing oil production IoT optimization security analytics
Production optimization Pipeline monitoring Trade monitoring, Real-time anomaly detection Fast marketing and optimization multi-channel engagement Integrate exploration Preventive maintenance Predictive analytics and seismic data Retail mobile applications Develop new products and Smart grids and microgrids Industrial safety monitor acceptance of rates Minimize lease Vendor management - operating expenses Grid operations, field service construction, transportation, Environment health and safety Predictive energy trading truck and delivery Decline curve analysis Asset performance Deep customer insights as a service optimization
Faster innovation Improved outcomes Optimizing supply- Predictive analytics Improved customer for revenue and increased chain with machine transforms safety engagement with growth revenue learning and security machine learning Security controls to leverage Actionable threat Risk and fraud Compliance Identity and access all data intelligence analysis management management for analytics
Firewall/network logs Firewall/network logs Firewall/network logs Firewall/network logs Files Apps Network flows Web/app logs Web Tables Data access layers Authentications Social media content Applications Clusters Devices Reports OS Dashboards Notebooks
Intrusion detection Security Fraud detection Security compliance Fine-grained data and predictive intelligence and prevention reporting analytics security analytics Prevention of DDoS attacks Real-time data correlation e-Tailing Ad-hoc/historic incident Role-based access controls reports Threat classifications Anomaly detection Inventory monitoring Auditing and governance SOC/NOC dashboards Data loss/anomaly detection Security context, enrichment Social media monitoring File integrity monitoring in streaming Deep OS auditing Offence scoring, prioritization Phishing scams Row level and column level Cybermetrics and changing Data loss detection in IoT access permissions use patterns Security orchestration Piracy protection User behavior analytics
Prevent complex Faster innovation Risk management Transform security Limit malicious threats with for threat with machine with improved insiders to machine learning prevention learning visibility transform growth
Vision Speech Language Vision Speech Language Accelerate time to value Innovate with AI everywhere – with agile tools and services in the cloud, at edge and on-premises
Pretrained AI Powerful Comprehensive Cloud Edge On-premises services tools platform
Use any language, any development Benefit from industry-leading security, privacy, tool and any framework compliance, transparency, and AI ethics standards
>90% of Fortune 500 companies use Microsoft Cloud It’s all on Microsoft Azure
© Microsoft Corporation Microsoft Azure
Productive Hybrid Intelligent Trusted Accelerate time to market Optimize your infrastructure Innovate at scale Develop with confidence
© Microsoft Corporation © Microsoft Corporation Data integration System integration BI and analytics
© Microsoft Corporation
Announcing ONNX Runtime open source
Create Deploy
Frameworks Azure
Azure Machine Learning services
Ubuntu VM
Windows Server 2019 VM
ONNX Model Devices
Services Windows devices
Azure Custom Vision Service Other devices (iOS, etc.) To empower data science and development teams
+ Azure Machine Learning Azure Databricks Python-based machine learning service Apache Spark-based big-data service
Develop models faster with automated machine learning Prepare data clean data at massive scale Use any Python environment and ML frameworks Enable collaboration between data scientists and data engineers Manage models across the cloud and the edge. Access machine learning optimized clusters Deploy Azure ML models at scale
Azure Machine Learning service What to use when?
Customer use case Data Prep Build & Train Manage and Deploy +
Big Data/Apache Spark Azure Databricks Azure Databricks + Azure ML service Azure ML service
(Apache Spark Dataframes, Datasets, Delta, (Spark MLlib and OSS frameworks + (containerize, deploy, inference and Pandas, NumPy etc.) Automated ML, Model Registry) monitor)
Data Science Pandas, NumPy etc. Azure ML service Azure ML service
(OSS frameworks, Hyperdrive, Pipelines, (containerize, deploy, inference and Automated ML, Model Registry) monitor) DevOps loop for data science
Prepare Experiment Deploy
…
Prepare Build model Train & Register and Build Deploy Service Data (your favorite IDE) Test Model Manage Model Image Monitor Model DevOps loop for data science
Train & Prepare Test Model
Register and … Manage Model
Build model (your favorite IDE)
Prepare Data Deploy Service Monitor Model Build Image Model management in Azure Machine Learning
Create/retrain model
Register model Monitor
Deploy Create scoring image Cloud Light Heavy files and Edge Edge dependencies
Create and register image Experimentation
Leverage service-side capture of run metrics, Manage training jobs locally, scaled-up or output logs and models scaled-out
80% 75%
95 90% % 85%
Use leaderboards, side by side run comparison Conduct a hyperparameter search on and model selection traditional ML or DNN