book

The Machine Learning Solutions Architect Handbook - Second Edition

Name: The Machine Learning Solutions Architect Handbook - Second Edition
Author: David Ping
ISBN: 9781805122500

by David Ping

April 2024

Intermediate to advanced

602 pages

16h 37m

English

Packt Publishing

Read now

Unlock full access

Preface
Who this book is forWhat this book coversTo get the most out of this bookGet in touch
Navigating the ML Lifecycle with ML Solutions Architecture
ML versus traditional softwareML lifecycleBusiness problem understanding and ML problem framingData understanding and data preparationModel training and evaluationModel deploymentModel monitoringBusiness metric trackingML challengesML solutions architectureBusiness understanding and ML transformationIdentification and verification of ML techniques System architecture design and implementationML platform workflow automation Security and complianceSummary
Exploring ML Business Use Cases
ML use cases in financial servicesCapital market front officeSales trading and researchInvestment bankingWealth managementCapital market back office operationsNet Asset Value reviewPost-trade settlement failure predictionRisk management and fraudAnti-money launderingTrade surveillanceCredit riskInsuranceInsurance underwritingInsurance claim managementML use cases in media and entertainmentContent development and productionContent management and discoveryContent distribution and customer engagementML use cases in healthcare and life sciencesMedical imaging analysisDrug discoveryHealthcare data managementML use cases in manufacturingEngineering and product designManufacturing operations – product quality and yieldManufacturing operations – machine maintenanceML use cases in retailProduct search and discoveryTargeted marketingSentiment analysisProduct demand forecastingML use cases in the automotive industryAutonomous vehiclesPerception and localizationDecision and planningControlAdvanced driver assistance systems (ADAS)Summary
Exploring ML Algorithms
Technical requirementsHow machines learnOverview of ML algorithmsConsideration for choosing ML algorithmsAlgorithms for classification and regression problemsLinear regression algorithmsLogistic regression algorithmsDecision tree algorithmsRandom forest algorithmGradient boosting machine and XGBoost algorithmsK-nearest neighbor algorithmMulti-layer perceptron (MLP) networksAlgorithms for clusteringAlgorithms for time series analysisARIMA algorithmDeepAR algorithmAlgorithms for recommendationCollaborative filtering algorithmMulti-armed bandit/contextual bandit algorithmAlgorithms for computer vision problemsConvolutional neural networksResNetAlgorithms for natural language processing (NLP) problemsWord2VecBERTGenerative AI algorithmsGenerative adversarial networkGenerative pre-trained transformer (GPT)Large Language ModelDiffusion modelHands-on exerciseProblem statementDataset descriptionSetting up a Jupyter Notebook environmentRunning the exerciseSummary
Data Management for ML
Technical requirementsData management considerations for MLData management architecture for MLData storage and managementAWS Lake FormationData ingestionKinesis FirehoseAWS GlueAWS LambdaData catalogingAWS Glue Data CatalogCustom data catalog solutionData processingML data versioningS3 partitionsVersioned S3 bucketsPurpose-built data version toolsML feature storesData serving for client consumptionConsumption via APIConsumption via data copySpecial databases for ML Vector databasesGraph databasesData pipelinesAuthentication and authorizationData governanceData lineageOther data governance measuresHands-on exercise – data management for MLCreating a data lake using Lake FormationCreating a data ingestion pipelineCreating a Glue Data CatalogDiscovering and querying data in the data lakeCreating an Amazon Glue ETL job to process data for MLBuilding a data pipeline using Glue workflowsSummary
Exploring Open-Source ML Libraries
Technical requirementsCore features of open-source ML librariesUnderstanding the scikit-learn ML libraryInstalling scikit-learnCore components of scikit-learnUnderstanding the Apache Spark ML libraryInstalling Spark MLCore components of the Spark ML libraryUnderstanding the TensorFlow deep learning libraryInstalling TensorFlowCore components of TensorFlowHands-on exercise – training a TensorFlow modelUnderstanding the PyTorch deep learning libraryInstalling PyTorchCore components of PyTorchHands-on exercise – building and training a PyTorch modelHow to choose between TensorFlow and PyTorchSummary
Kubernetes Container Orchestration Infrastructure Management
Technical requirementsIntroduction to containersOverview of Kubernetes and its core conceptsNamespacesPodsDeploymentKubernetes JobKubernetes custom resources and operatorsServicesNetworking on KubernetesSecurity and access managementAPI authentication and authorizationHands-on – creating a Kubernetes infrastructure on AWSProblem statementLab instructionSummary
Open-Source ML Platforms
Core components of an ML platformOpen-source technologies for building ML platformsImplementing a data science environmentBuilding a model training environmentRegistering models with a model registryServing models using model serving servicesThe Gunicorn and Flask inference engineThe TensorFlow Serving frameworkThe TorchServe serving frameworkKFServing frameworkSeldon CoreTriton Inference ServerMonitoring models in productionManaging ML featuresAutomating ML pipeline workflowsApache AirflowKubeflow PipelinesDesigning an end-to-end ML platformML platform-based strategyML component-based strategySummary
Building a Data Science Environment Using AWS ML Services
Technical requirementsSageMaker overviewData science environment architecture using SageMakerOnboarding SageMaker usersLaunching Studio applicationsPreparing dataPreparing data interactively with SageMaker Data WranglerPreparing data at scale interactivelyProcessing data as separate jobsCreating, storing, and sharing featuresTraining ML modelsTuning ML modelsDeploying ML models for testingBest practices for building a data science environmentHands-on exercise – building a data science environment using AWS servicesProblem statementDataset descriptionLab instructionsSetting up SageMaker StudioLaunching a JupyterLab notebookTraining the BERT model in the Jupyter notebookTraining the BERT model with the SageMaker Training serviceDeploying the modelBuilding ML models with SageMaker CanvasSummary
Designing an Enterprise ML Architecture with AWS ML Services
Technical requirementsKey considerations for ML platformsThe personas of ML platforms and their requirementsML platform buildersPlatform users and operatorsCommon workflow of an ML initiativePlatform requirements for the different personasKey requirements for an enterprise ML platformEnterprise ML architecture pattern overviewModel training environmentModel training engine using SageMakerAutomation supportModel training lifecycle managementModel hosting environmentInference enginesAuthentication and security controlMonitoring and loggingAdopting MLOps for ML workflowsComponents of the MLOps architectureMonitoring and loggingModel training monitoringModel endpoint monitoringML pipeline monitoringService provisioning managementBest practices in building and operating an ML platformML platform project execution best practicesML platform design and implementation best practicesPlatform use and operations best practicesSummary

Advanced ML Engineering
Technical requirementsTraining large-scale models with distributed trainingDistributed model training using data parallelismParameter server overviewAllReduce overviewDistributed model training using model parallelismNaïve model parallelism overviewTensor parallelism/tensor slicing overviewImplementing model-parallel trainingAchieving low-latency model inferenceHow model inference works and opportunities for optimizationHardware accelerationCentral processing units (CPUs)Graphics processing units (GPUs)Application-specific integrated circuitModel optimizationQuantizationPruning (also known as sparsity)Graph and operator optimizationGraph optimizationOperator optimizationModel compilersTensorFlow XLAPyTorch GlowApache TVMAmazon SageMaker NeoInference engine optimizationInference batchingEnabling parallel serving sessionsPicking a communication protocolInference in large language modelsText Generation Inference (TGI)DeepSpeed-InferenceFastTransformerHands-on lab – running distributed model training with PyTorchProblem statementDataset descriptionModifying the training scriptModifying and running the launcher notebookSummary
Building ML Solutions with AWS AI Services
Technical requirementsWhat are AI services?Overview of AWS AI servicesAmazon ComprehendAmazon TextractAmazon RekognitionAmazon TranscribeAmazon PersonalizeAmazon Lex V2Amazon KendraAmazon QEvaluating AWS AI services for ML use casesBuilding intelligent solutions with AI servicesAutomating loan document verification and data extractionLoan document classification workflowLoan data processing flowMedia processing and analysis workflowE-commerce product recommendationCustomer self-service automation with intelligent searchDesigning an MLOps architecture for AI servicesAWS account setup strategy for AI services and MLOpsCode promotion across environmentsMonitoring operational metrics for AI servicesHands-on lab – running ML tasks using AI servicesSummary
AI Risk Management
Understanding AI risk scenariosThe regulatory landscape around AI risk managementUnderstanding AI risk managementGovernance oversight principlesAI risk management frameworkApplying risk management across the AI lifecycleBusiness problem identification and definitionData acquisition and managementRisk considerationsRisk mitigations Experimentation and model developmentRisk considerationsRisk mitigationsAI system deployment and operationsRisk considerationsRisk mitigationsDesigning ML platforms with governance and risk management considerationsData and model documentationLineage and reproducibilityObservability and auditingScalability and performanceData qualitySummary
Bias, Explainability, Privacy, and Adversarial Attacks
Understanding biasUnderstanding ML explainabilityLIMESHAPUnderstanding security and privacy-preserving MLDifferential privacyUnderstanding adversarial attacksEvasion attacksPGD attacksHopSkipJump attacksData poisoning attacksClean-label backdoor attackModel extraction attackAttacks against generative AI modelsDefense against adversarial attacksRobustness-based methodsDetector-based methodOpen-source tools for adversarial attacks and defensesHands-on lab – detecting bias, explaining models, training privacy-preserving mode, and simulating adversarial attackProblem statementDetecting bias in the training datasetExplaining feature importance for a trained modelTraining privacy-preserving modelsSimulate a clean-label backdoor attackSummary
Charting the Course of Your ML Journey
ML adoption stagesExploring AI/MLDisjointed AI/MLIntegrated AI/MLAdvanced AI/MLAI/ML maturity and assessmentTechnical maturityBusiness maturityGovernance maturityOrganization and talent maturityMaturity assessment and improvement processAI/ML operating modelsCentralized modelDecentralized modelHub and spoke modelSolving ML journey challengesDeveloping the AI vision and strategyGetting started with the first AI/ML initiativeSolving scaling challenges with AI/ML adoption Solving ML use case scaling challengesSolving technology scaling challengesSolving governance scaling challengesSummary
Navigating the Generative AI Project Lifecycle
The advancement and economic impact of generative AIWhat industries are doing with generative AIFinancial servicesHealthcare and life sciencesMedia and entertainmentAutomotive and manufacturingThe lifecycle of a generative AI project and the core technologiesBusiness use case selectionFM selection and evaluationInitial screening via manual assessmentAutomated model evaluationHuman evaluationAssessing AI risks for FMsOther evaluation considerationBuilding FMs from scratch via pre-trainingAdaptation and customizationDomain adaptation pre-trainingFine-tuningReinforcement learning from human feedbackPrompt engineeringModel management and deploymentThe limitations, risks, and challenges of adopting generative AI Summary
Designing Generative AI Platforms and Solutions
Operational considerations for generative AI platforms and solutionsNew generative AI workflow and processesNew technology componentsNew rolesExploring generative AI platformsThe prompt management componentFM benchmark workbenchSupervised fine-tuning and RLHFFM monitoringThe retrieval-augmented generation patternOpen-source frameworks for RAGLangChainLlamaIndexEvaluating a RAG pipelineAdvanced RAG patternsDesigning a RAG architecture on AWSChoosing an LLM adaptation methodResponse qualityCost of the adaptationImplementation complexityBringing it all togetherConsiderations for deploying generative AI applications in productionModel readiness Decision-making workflowResponsible AI assessmentGuardrails in production environmentsExternal knowledge change managementPractical generative AI business solutionsGenerative AI-powered semantic search engineFinancial data analysis and research workflowClinical trial recruiting workflowMedia entertainment content creation workflowCar design workflowContact center customer service operationAre we close to having artificial general intelligence?The symbolic approachThe connectionist/neural network approachThe neural-symbolic approachSummary
Other Books You May Enjoy
Index

Content preview from The Machine Learning Solutions Architect Handbook - Second Edition

16 Designing Generative AI Platforms and Solutions

Deploying generative AI at scale in an enterprise introduces new complexities around infrastructure, tooling, and operational processes required to harness its potential while managing risks. This chapter explores the essential components for building robust generative AI platforms and examines Retrieval-Augmented Generation (RAG), an effective architecture pattern for generative applications. Additionally, we highlight near-term generative AI solution opportunities ripe for business adoption across industries. With the right platform foundations and a pragmatic approach focused on delivering tangible value, enterprises can start realizing benefits from generative AI today while paving the way ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

The Machine Learning Solutions Architect Handbook

Publisher Resources

ISBN: 9781805122500

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

The Machine Learning Solutions Architect Handbook - Second Edition

by David Ping

16

Designing Generative AI Platforms and Solutions

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.