Transcription
Nishant ThackerMicrosoft Confidential
3.9 TDecisionsupportGlobal business value derivedfrom AI in 2022 will reachDecisionautomationSmartproducts 3.9TVirtualagents“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 CapabilitiesBASICADVANCED
ductProduct RecommendationRecommendationLead ScoringIntelligent chatbotsFinancial ForecastingPredictivePredictive MaintenanceMaintenanceEmployee InsightsCustomer InsightsSales InsightsVirtual AssistantsCash Flow ForecastingDemandDemand ForecastingForecastingHR InsightsChurn AnalyticsDynamic PricingWaiting line optimizationRisk ManagementQuality AssuranceResource Planning
Product recommendationPredictive maintenanceDemand forecastingThe average size of a single cart hasdecreasedUnplanned downtime results in costoverrunsSolar energy production is inconsistentProvide personalized digital content toshoppersPredict when maintenance should beperformedAlign energy supply with the optimalmarketsIncrease cart sizeMinimize downtimeMaximize revenueASOS delivers 15.4 million personalizedexperiences with 33 orders per secondHybrid solution predicts onboard waterusage, saving 200k/ship/yearDistributed power generation increasesrevenue by over 100 million
Serving business users and end users with intelligent and dynamicapplicationsBuild a unified and usabledata pipelineTrain ML and DL models toderive insightsOperationalize models anddistribute insights at scale
Complexity of solutionsData silosAndmostaren’t satisfiedwith their current solutions Incongruent data typesManyoptionsin the marketplace46%Are satisfied with ease ofuse of analytics software“What it Takes to be Data-Driven: Technologies and Best Practices for Becoming a Smarter Organization”, TDWI, 2017.21%Are satisfied with access to semistructured and unstructured dataDifficult to scale effectivelyPerformance constraints28%Are satisfied with ability to scale tohandle unexpected requirements
DataAI
DataAIData Modernization on-premisesAI apps & agentsData modernization to AzureKnowledge miningGlobally distributed dataCloud Scale Analytics
Machine Learning on AzureDomain specific pretrained modelsTo reduce time to market VisionSpeechLanguageSearchPyCharmJupyterVisual Studio CodeCommand linePytorchTensorFlowScikit-LearnOnnxFamiliar Data Science toolsTo simplify model developmentPopular frameworksTo build advanced deep learning solutionsProductive servicesTo empower data science and development teamsAzureDatabricksAzure MachineLearningMachineLearning VMsCPUGPUFPGAPowerful infrastructureTo accelerate deep learningFrom the Intelligent Cloud to the Intelligent Edge
Infuse apps with powerful, pre-trained AI modelsCustomize easily and tailor to your needsLanguageVisionText Analytics Spell Check Language Understanding Text Translation QnA MakerComputer Vision Video Indexer Face Content Moderator SpeechBingSearchBig Web Search Video Search Image Search Visual Search Entity Search News Search AutosuggestSpeech to Text Text to Speech Speech Translation Speaker Recognition
Empower data science and development teamsIntegrated data science & data engineering teamsIndividual data scientistsDesktop solutions not adequateDesktop solutions adequateNeed a unified big data & machine learning solutionNeed cloud for sporadic compute needs Azure DatabricksAzure Machine LearningserviceMachine Learning VMs
Fast, easy, and collaborative Apache Spark -based analytics platformIncrease productivityBuilt with your needs in mindOptimized Apache Spark environmnetCollaborative workspaceBuild on a secure, trusted cloudIntegration with Azure data servicesAutoscale and autoterminateOptimized for distributed processingScale without limitsSupport for multiple languages and librariesSeamlessly integrated with the Azure Portfolio
Bring AI to everyone with an end-to-end, scalable, trusted platformBoost your data science productivityBuilt with your needs in mindAutomated machine learningManaged computeIncrease your rate of experimentationDevOps for machine learningSimple deploymentTool agnostic Python SDKDeploy and manage your models everywhereSupport for open source frameworksSeamlessly integrated with the Azure Portfolio
Accelerate deep learningCPUsGPUsFPGAsGeneral purpose machinelearningDeep learningSpecialized hardwareaccelerated deep learningD, F, L, M, H SeriesOptimized for flexibilityN SeriesProject BrainwaveOptimized for performance
From the Intelligent Cloud to the Intelligent EdgeTrain and deployTrain and deployTrack models in productionCapture model telemetryRetrain modelsDeploy
Quickly and easily build, train, deploy and manage your modelsDomain Specific Pretrained ModelsTo reduce time to market VisionSpeechLanguageSearchPyCharmJupyterVisual Studio CodeCommand lineFamiliar Data Science toolsTo simplify model developmentPopular FrameworksTo build machine learning and deep learning ve ServicesTo empower data science and development teamsAzureDatabricksAzure MachineLearningMachineLearning VMsCPUGPUFPGAPowerful InfrastructureTo accelerate deep learningFrom the Intelligent Cloud to the Intelligent Edge
IngestStoreServePrep and trainModel ServingAzure KubernetesserviceAzure EventHubsStreaming dataOperationalDatabasesAppsCosmos DB, SQL DBAzure Databricks Azure MachineLearning serviceAd-hoc AnalysisBatch dataAzure Data FactoryAzure Data Lake StorageAzure SQL datawarehousePower BIAzure analysisservices
PREP & TRAINABCCollect and prepare dataAzure Data FactoryAzure DatabricksTrain and evaluate modelAzure DatabricksOperationalize and manageAzure ML ServicesAzure Databricks
Azure DataFactoryAzure BlobStorageIngestAzureDatabricksStoreUnderstand and transformConnect to datafrom any sourceLeverage best-in-classanalytics capabilitiesScalewithout 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
ToolsAzure MLserviceAzureDatabricksAutomated MLAzure MLserviceMachine learningAzure MLserviceInfrastructureAutomated ML capabilitiesAzureDatabricksScale out clustersSimplify modeldevelopmentScale compute resourcesto meet your needsQuickly determine theright 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
AzureML serviceAzureML serviceAKSACIContainersDockerTrain and evaluate modelsModel MGMT, experimentation,and run historyIoT edgeBring modelsto life quicklyProactively managemodel performanceDeploy modelscloser 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 ScienceVMsPyTorchKerasMS CognitiveToolkitAI Ready Virtual MachinesAzure MLservice SDKTensorFlowNotebooks, IDEs, Command LineAML ComputeScale out clustersStreamlineAI development effortsScale computeresources in any environmentQuickly evaluateand 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 MLToolsFrameworksInfrastructureUse HorovodEstimator via a nativeruntime to enable build deep learningmodels with a few lines of codeFull Python and Scala support fortransfer learning on imagesLeverage powerful GPU-enabled VMspre-configured for deep neuralnetwork trainingLoad images natively in SparkDataFrames to automatically decodethem for manipulation at scale withdistributed DNN training on SparkSeamlessly use TensorFlow, MicrosoftCognitive Toolkit, Caffe2, Keras, andmoreSimultaneously collaborate withinnotebooks environments tostreamline model developmentUse built-in hyperparameter tuningvia Spark MLLib to quickly optimizethe modelAutomatically store metadata in AzureDatabase with geo-replication forfault toleranceImprove performance 10x-100x overtraditional Spark deployments with anoptimized environment
Azure Databricks Remote support for other IDEs outside of native notebooksMLFlow for better DevOps with Azure Databricks and other ML pipelinesAzure Machine Learning Python SDK support for popular IDEs & notebooks, including Azure DatabricksAzure Machine Learning managed compute capabilitiesIntroduce new models for FPGA scoringRobust ONNX support - runtime engine in AML, model operationalization in SQL ServerAutomated machine learningDeploy and manage models to IoT edgeExtend Machine Learning services to SQL DB
AZURE ML SERVICEAZURE DATABRICKSBring AI to the edgeAccelerate processing with the fastest Apache Spark engineIncrease your rate of experimentationIntegrate natively with Azure servicesDeploy and manage your models everywhereAccess enterprise-grade Azure securityLeverage your favorite deep learning frameworksTensorFlowMS Cognitive ToolkitPyTorchScikit-LearnONNXCaffe2MXNetChainer
Get started quickly withDeveloper AI Microsoft Corporation
Azure bot service cognitive servicesBing searchTextStartState 1State 2Dynamics365SpeechState 3Bot frameworkSDK dialogmanagerOn-premisesSpeechIFTTTQnA MakerImageTranslator Microsoft CorporationDispatchLanguageunderstanding
Cognitive servicesAzure searchBot servicesUse pre-built AI servicesto solve business problemsGet up andrunning quicklySpeed development with a purpose-builtenvironment for bot creationMap complexinformation and dataReduce complexity witha fully-managed serviceInfuse intelligence into your bot usingcognitive servicesAllow your apps toprocess natural languageUse artificial intelligenceto extract insightsIntegrate across multiplechannels to reach more customersCreate a seamless developer experience across desktop, cloud, or at the edge using Visual Studio AI Tools Microsoft Corporation
Conversational agentsIntelligent appsBusiness processesTransform your engagements withLeverage AI to create the futureTransform critical businesscustomers and employeesof business applicationsprocesses with AI95% Microsoft CorporationOf customer interactionspowered by AI bots by 202575%Applications to includeAI by the end of this year85%Of enterprises usingAI by 2020
DronesSmart factoryVoice-activated speakersSmart kiosksSmart kitchenSmart camerasSmart bath Microsoft Corporation
Empowering people with tools that amplify human capabilityEmploymentModern lifeHuman connectionFacilitating developmentof professional skillsDelivering personalized experiencesto improve independenceProviding equal access toinformation and opportunityHow Microsoftis helping thevisually-impaired Microsoft CorporationHow Helpictois helpingnon-verbal childrenHow RITis helping thehearing-impaired
Effective customerengagementDecision servicesmanagementRisk and revenuemanagementRisk and compliancemanagementRecommendationengineCustomer profilesCustomer segmentationTransaction dataCRMClickstream dataCredit historyCRM dataDemographicsCreditProductsTransactional dataCredit dataPurchasing historyRiskServicesLTVMarket dataTrendsMerchant recordsCustomer service dataLoyaltyCustomeranalyticsCustomer 360degree evaluationCustomer segmentationReduced customer churnUnderwriting, servicing anddelinquency handlingProducts and servicesFinancialmodelingCommercial/retail banking,securities, trading andinvestment modelsDecision science, simulationsand forecastingInvestment recommendationsRisk, fraud,threat e anomaly detectionEnterprise DataHubRecommendation engineCard monitoring and frauddetectionRegulatory andcompliance analysisPredictive analytics andtargeted advertisingSecurity threat identificationCredit risk managementRisk aggregationAutomated credit analyticsFast marketing and multichannel engagementCustomer sentiment analysisInsights for new productsFaster innovationfor a bettercustomer experienceImproved consumeroutcomes andincreased revenueEnhanced customerexperience withmachine learningTransform growthwith predictiveanalyticsImproved customerengagement withmachine learning
DNAsequencesReal worldanalyticsImagedeep learningSensordataSocial datalisteningFAST-QHL7/CCDMRIReadingsSocial mediaBAM837X-RAYTime seriesAdverse eventsSAMPharmacyCTEvent Genomics andprecision medicineClinical andclaims dataIoT deviceanalyticsSocialanalyticsGPU imageprocessingSingle cell sequencingClaims data warehouseGraphic intensive workloadsBiomarker, genetic, variantand population analyticsReadmission predictionsDeep learning usingTensor FlowADAM and HAIL on DatabricksEfficacy and comparativeanalyticsPattern recognitionAggregation ofstreaming eventsReal-time patient feedbackvia topic modellingPredictive maintenanceAnalytics acrosspublication dataAnomaly detectionPrescription adherenceMarket access analysisFaster innovationfor drugdevelopmentImproved outcomesand increasedrevenueDiagnosticsleveragingmachine learningPredictive analyticstransforms qualityof careImproved patientcommunicationsand feedback
PersonalizedrecommendationsEffectivecustomer nConsumer engagementanalysisCustomer profilesCustomer profilesConsumption logsTransactionsContent metadataViewing historyOnline activityClickstream and devicesSubscriptionsRatingsOnline activityContent distributionDemographicsCommentsContent sourcesServices dataMarketing campaignresponsesCredit dataSocial media er churnpreventionPersonalized viewing andengagement experienceQuality of service andoperational efficiencyClick-path optimizationMarket basket analysisNext best content analysisCustomer behavior analysisImproved real timead targetingClick-through analysisAd effectivenessPredict audience interestsDemand-elasticityContent monetizationNetwork performanceand optimizationSocial network analysisFraud detectionInformation-as-a-serviceHigh value user engagementPricing predictionsPromotion eventstime-series analysisNielsen ratings andprojectionsMulti-channel marketingattributionMobile spatial analyticsFaster innovationfor customerexperienceImproved consumeroutcomes andincreased revenueEnhance userexperience withmachine learningPredictiveanalyticstransforms growthImproved consumerengagement withmachine learning
RecommendationengineEffective locationConsumerengagementCustomer profilesShopping historyDemand plansDemographicsHistorical sales dataShopping historyOnline activityForecastsBuyer perceptionPrice schedulingOnline activityFloor plansSales historyConsumer researchSegment level price changesSocial network analysisApp dataTrendsMarket/competitive analysisLocal events/weatherpatternsNext best andpersonalized offersCustomer 360/consumerpersonalizationRight product, promotion,at right timeMulti-channel promotionFaster innovationfor customerexperienceStore design andergonomicsPath to purchaseIn-store experienceWorkforce and manpoweroptimizationImproved consumeroutcomes andincreased revenueData-driven stock,inventory, orderingPredict inventory positionsand distributionAssortmentoptimizationReal-time pricingoptimizationEconomic modellingDemand-elasticityFraud detectionOptimization for foot traffic,Online interactionsPersonal pricing schemesMarket basket analysisFlat and declining categoriesOmni-channelshoppingexperience withmachine learningPredictiveanalyticstransforms growthPromotion eventsMulti-channel engagementImproved consumerengagement withmachine learning
Effective customerengagementRecommendationengineRisk and fraudanalysisCampaign reportinganalyticsBrand promotion andcustomer experienceCustomer profilesCustomer segmentationTransaction dataCRMSocial mediaOnline historyCRM dataDemographicsMerchant recordsOnline historyTransaction dataCredit dataPurchasing historyProductsCustomer service dataLoyaltyMarket dataTrendsServicesMarketing dataCustomer valueanalyticsNext best andpersonalized offersCustomer 360, segmentationaggregation and attributionRight product, promotion,at right timeAudience modelling/indexreportReal time ad bidding platformReduce customer churnInsights for new productsRisk and fraudmanagementReal-time anomaly detectionFraud preventionPersonalized ad targetingCustomer spend andrisk analysisAd performance reportingData relationship mapsSales and campaignoptimizationSentimentanalysisOptimizing return on adspend and ad placementOpinion mining/socialmedia analysisMulti-channel promotionDeeper customer insightsIdeal customer traitsCustomer loyalty programsOptimized ad placementShopping cart analysisHistorical bid opportunityas a serviceFaster innovationfor customergrowthImproved outcomesand increasedrevenueRisk managementwith machinelearningPredictiveanalyticstransforms growthImproved customerengagement withmachine learning
Upstream optimization,maximize well lifeGrid operations, assetinventory onRecommendationsengineField dataSensor stream dataTransaction dataSensor stream dataClickstream dataAsset dataUAVs entory dataPurchasing historyRetail dataServicesProduction dataProduction dataTrendsGrid production dataMarket dataRefinery tuning parametersCompetitive dataDemographicsDigital oil field/oil productionIndustrialIoTProduction optimizationPipeline monitoringIntegrate explorationand seismic dataPreventive maintenanceMinimize leaseoperating expensesDecline curve analysisFaster innovationfor revenuegrowthSmart grids and microgridsGrid operations, field serviceAsset performanceas a serviceImproved outcomesand increasedrevenueSupply-chainoptimizationTrade monitoring,optimizationRetail mobile applicationsVendor management construction, transportation,truck and deliveryoptimizationOptimizing supplychain with machinelearningSafety andsecurityReal-time anomaly detectionPredictive analyticsIndustrial safetyEnvironment health and safetySales and marketinganalyticsFast marketing andmulti-channel engagementDevelop new products andmonitor acceptance of ratesPredictive energy tradingDeep customer insightsPredictive analyticstransforms safetyand securityImproved customerengagement withmachine learning
Security controls to leverageall dataActionable threatintelligenceRisk and fraudanalysisCompliancemanagementIdentity and accessmanagement for analyticsFirewall/network logsFirewall/network logsFirewall/network logsFirewall/network logsFilesAppsNetwork flowsWeb/app logsWebTablesData access layersAuthenticationsSocial media rdsNotebooksIntrusion detectionand predictiveanalyticsSecurityintelligenceFraud detectionand preventionPrevention of DDoS attacksReal-time data correlatione-TailingThreat classificationsAnomaly detectionInventory monitoringData loss/anomaly detectionin streamingSecurity context, enrichmentSocial media monitoringOffence scoring, prioritizationPhishing scamsSecurity orchestrationPiracy protectionCybermetrics and changinguse patternsPrevent complexthreats withmachine learningFaster innovationfor threatpreventionRisk managementwith machinelearningSecurity compliancereportingAd-hoc/historic incidentreportsSOC/NOC dashboardsDeep OS auditingData loss detection in IoTFine-grained dataanalytics securityRole-based access controlsAuditing and governanceFile integrity monitoringRow level and column levelaccess permissionsUser behavior analyticsTransform securitywith improvedvisibilityLimit maliciousinsiders totransform growth
VisionSpeechLanguage
VisionSpeechLanguage
Accelerate time to valuewith agile tools and servicesPretrained AIservicesPowerfultoolsComprehensiveplatformUse any language, any developmenttool and any frameworkInnovate with AI everywhere –in the cloud, at edge and on-premisesCloudEdgeOn-premisesBenefit from industry-leading security, privacy,compliance, transparency, and AI ethics standards 90% of Fortune 500 companiesuse Microsoft Cloud
It’s all on Microsoft Azure Microsoft Corporation
Microsoft AzureProductiveHybridIntelligentTrustedAccelerate time to marketOptimize your infrastructureInnovate at scaleDevelop with confidence Microsoft Corporation
Microsoft Corporation
Data integration Microsoft CorporationSystem integrationBI and analytics
Announcing ONNX Runtime open sourceCreateDeployFrameworksAzureAzure Machine Learning servicesUbuntu VMWindows Server 2019 VMONNX ModelDevicesServicesAzure Custom Vision ServiceWindows devicesOther devices (iOS, etc.)
To empower data science and development teamsAzure Machine Learning Azure DatabricksPython-based machine learning serviceApache Spark-based big-data serviceDevelop models faster with automated machine learningPrepare data clean data at massive scaleEnable collaboration between data scientists and data engineersAccess machine learning optimized clustersUse any Python environment and ML frameworksManage models across the cloud and the edge.
Deploy Azure ML models at scaleAzure Machine Learning service
What to use when?Big Data/Apache SparkData ScienceData PrepBuild & TrainManage and DeployAzure DatabricksAzure Databricks Azure ML serviceAzure ML service(Apache Spark Dataframes, Datasets, Delta,Pandas, NumPy etc.)(Spark MLlib and OSS frameworks Automated ML, Model Registry)(containerize, deploy, inference andmonitor)Azure ML serviceAzure ML service(OSS frameworks, Hyperdrive, Pipelines,Automated ML, Model Registry)(containerize, deploy, inference andmonitor) Customer use casePandas, NumPy etc.
DevOps loop for data sciencePrepareDeployExperiment PrepareDataBuild model(your favorite IDE)Train &Test ModelRegister andManage ModelBuildImageDeploy ServiceMonitor Model
DevOps loop for data scienceTrain &Test ModelPrepareRegister andManage Model Build model(your favorite IDE)PrepareDataDeploy ServiceMonitor ModelBuildImage
Model management in Azure Machine LearningCreate/retrain modelRegister modelMonitorDeployimageCloudLightEdgeCreate scoringfiles anddependenciesHeavyEdgeCreate and register image
ExperimentationLeverage service-side capture of run metrics,output logs and modelsManage training jobs locally, scaled-up orscaled-out80%75%90%95%85%Use leaderboards, side by side run comparisonand model selectionConduct a hyperparameter search ontraditional ML or DNN
Popular frameworks To build advanced deep learning solutions . Speech Language Vision Search Productive services To empower data science and development teams Powerful infrastructure To accelerate deep learning Scikit-Learn PyCharm Jupyter Familiar Data Science tools . Customer service data Customer
