AI-Driven Development Lifecycle

Amazon Web Services 1 AI - D riven Development Lifecycle (AI - DLC) Method DefinitionRaja SP, Amazon Web ServicesI. CONTEXTThe evolution of software engineering has been a continuousquest to enable developers to focus on solving complex problemsby abstracting away lower-level, undifferentiated tasks. Fromearly machine code to high-level programming languages and theadoption of APIs and libraries, each step has significantly boosteddeveloper productivity. Now, the integration of Large LanguageModels has revolutionized how software is created, introducingconversational natural language interactions for tasks like codegeneration, bug detection, and test generation. This marks theAI-Assisted era, where AI enhances such fine-grained, specifictasks.As AI evolves, its applications are expanding beyond codegeneration to include requirements elaboration, planning, taskdecomposition, design, and real-time collaboration withdevelopers. This shift is kick-starting the AI-Driven era, where AIactively orchestrates the development process. But, existingsoftware development methods, designed for human-driven,long-running processes, are not fully aligned with AI’s speed,flexibility, and advanced capabilities (ex. agentic). Their relianceon manual workflows and rigid role definitions limits the abilityto fully leverage AI. Retrofitting AI into these methods not onlylimits its potential, but also reinforces outdated inefficiencies. Tofully leverage AI’s transformative power, SDLC methods need tobe reimagined. This reimagination requires AI to be a centralcollaborator, aligning workflows, roles, and iterations to enablefaster decision-making, seamless task execution, and continuousadaptability.This paper introduces and defines the AI-Driven DevelopmentLifecycle (AI-DLC), a reimagined, AI-native methodology designedto fully integrate the capabilities of AI, setting the foundations forthe next evolution in software engineering.II. KEY PRINCIPLESThe principles in this section form the foundation for defining AI-DLC, shaping its phases, roles, artifacts, and rituals. Theseassumptions are critical for validating the proposed method, asthey provide the underpinning rationale behind its design.1. REIMAGINE RATHER THAN RETROFITWe choose to reimagine a development method rather thankeeping the existing methods like SDLC or Agile (e.g., Scrum) andretrofitting AI into them. These traditional methods were builtfor longer iteration duration (months and weeks), which led torituals like daily standups and retrospectives. In contrast, properapplication of AI leads to rapid cycles, measured in hours or days.This needs continuous, real-time validation and feedbackmechanisms, rendering many of the traditional rituals lessrelevant. Would effort estimation (ex. story points) be as criticalif AI diminishes the boundaries between simple, medium andhard tasks? Would metrics like velocity be any relevant or shouldwe start replacing it with Business Value, as an example? Also, AIis increasingly evolving to automate manual practices includingplanning, task decomposition, requirements analysis, applicationof design techniques (ex. domain modelling), thereby shorteningthe number of phases it takes from moving intentions to code.These new dynamics warrant a reimagination based on firstprinciples thinking, rather than a retrofitting. We needautomobiles and not the faster horse chariots.2. REVERSE THE CONVERSATION DIRECTIONAI-DLC introduces a fundamental shift where AI initiates &directs the conversations with humans instead of humansinitiating the conversation with AI to complete their tasks. AIdrives workflows by breaking down high-level intents (ex.implementing a new business function) into actionable tasks,generating recommendations, and proposing trade-offs. Humansserve as approvers, validating, selecting options, and confirmingdecisions at critical junctures. This AI-driven approach allowsdevelopers to focus on high-value decision-making while AIhandles planning, task decomposition, and automation. Byreversing the traditional dynamic, AI-DLC ensures that humaninvolvement is purposeful, concentrating on oversight, riskmitigation, and strategic alignment, thereby enhancing bothvelocity and quality. An analogy to illustrate this is Google Maps:humans set the destination (the intent), and the system providesstep-by-step directions (AI’s task decomposition andrecommendations). Along the way, humans maintain oversightand moderate the journey as needed.3. INTEGRATION OF DESIGN TECHNIQUES INTO THE COREAgile frameworks like Scrum or Kanban leaves the designtechniques (ex. Domain Driven Design) out of scope andrecommends the teams to choose their own. This has left criticalwhitespaces that led to poor software quality overall. Softwarequality issues in US alone was estimated to cost $2.41 Trillion in2022 (study). Rather than decoupling design techniques out, AI-DLC will have them as its integral core. There will be differentflavors of AI-DLC for teams following Domain Driven Design(DDD), Behavior Driven Development (BDD) or Test-DrivenDevelopment (TDD) respectively. This paper discusses the DDD Amazon Web Services 2flavor of AI-DLC which will use DDD principles to break downsystems into independent, right-sized bounded contexts that canbe rapidly built in parallel. AI will inherently apply thesetechniques during planning and task decomposition, requiringdevelopers only to validate and adjust. This integration is key toenabling hourly or daily iteration cycles while eliminating manualheavy-lifting and maintaining software quality (the mantra of“Build Better Systems Faster”).4. ALIGN WITH AI CAPABILITYThis paper is optimistic about the future potential of AI but fullyrealistic about its current state. AI-DLC recognises that current AIis advancing but not yet reliable in autonomously translatinghigh-level intentions into executable code or independentlyoperating without human oversight, while also ensuringinterpretability and safety. At the same time, the AI-Assistedparadigm, where developers perform the majority of theintellectual heavy lifting with AI merely providing augmentation,fails to unlock the full potential of AI in development. AI-DLCadopts the AI-Driven paradigm, which balances humaninvolvement with the capabilities and limitations of current AI. Inthis, the developers retain ultimate responsibility for validation,decision-making, and oversight. This balance ensures that thestrengths of AI are leveraged effectively without compromisingthe critical safeguards provided by developer judgment.5. CATER TO BUILDING COMPLEX SYSTEMSAI-DLC caters to building systems that demand continuousfunctional adaptability, high architectural complexity, numeroustrade-offs management, scalability, integration andcustomization requirements. These necessitate the application ofadvanced design techniques, patterns, and best practices,typically involving multiple teams working cohesively withinlarge and/or regulated organizations. Simpler systems that can bedeveloped by non-developer personas that needed few or notrade-off management are outside the scope of AI-DLC and arebetter suited for low-code/no-code approaches.6. RETAIN WHAT ENHANCES HUMAN SYMBIOSISWhile reimagining the method, we will retain the artifacts andtouchpoints from the existing methods that are critical for humanvalidation and risk mitigation. For instance, user stories alignhumans’ and AI’s understanding of what needs to be built, actingas well-defined contracts. We will retain user stories as they arein the re-imagined method also. Another example is the RiskRegister that ensures AI-generated plans and codes comply withorganizational risk frameworks. These retained elements will beoptimized for real-time use, allowing rapid iterations withoutcompromising alignment or safety.7. FACILITATE TRANSITION THROUGH FAMILIARITYThe new method shall not demand extensive trainings and anyexisting practitioner should be able to orient and start practicingit in a single day. To support easier adoption via associativelearning, AI-DLC will preserve the underlying relationshipsbetween familiar terms in older methods while introducingmodernized terminology. For example, Sprints in Scrumrepresent iterative cycles for building and validating. But Sprintsare usually 4 to 6 weeks long in the pre-AI era. With AI-DLC, theiteration cycles will be continuous and in terms of hours or days.Therefore, we need to intentionally rename Sprints. AI-DLCrebrands Sprints as Bolts, emphasizing rapid, intense cycles thatdeliver unprecedented velocity.8. STREAMLINE RESPONSIBILITIES FOR EFFICIENCYBy leveraging AI’s ability to perform task decomposition, anddecision-making, developers will be empowered to transcendtraditional specialization silos such as infrastructure, front-end,back-end, DevOps, and security. This convergence ofresponsibilities reduces the need for multiple specialized roles,streamlining the development process. But Product Owners anddevelopers remain integral to the framework, retaining criticalresponsibilities for oversight, validation, and strategic decision-making. These roles ensure alignment with business objectives,maintain design quality, and maintain compliance with riskmanagement frameworks, preserving the balance betweenautomation and human accountability. In the method definition,we will stick to first principles, keeping the roles minimum, withadditional roles introduced only when critically necessary.9. MINIMISE STAGES, MAXIMISE FLOWThrough automation and convergence of responsibiliti4es, AI-DLC aims to minimize the handoffs and transitions, enablingcontinuous iterative flow. But human validation and decision-making remain critical to ensure that AI-generated code does notbecome rigid ('quick-cement') but stays adaptable for futureiterations. To address this, AI-DLC incorporates minimal butsufficient number of phases specifically designed for humanoversight at critical decision junctures. These validations act as aform of 'loss function’, by identifying and pruning wastefuldownstream efforts before they occur.10. NO HARD-WIRED, OPINIONATED SDLC WORKFLOWSAI-DLC avoids prescribing opinionated workflows for differentdevelopment pathways (such as new system development,refactoring, defect fixes, or microservice scaling). Instead, itadopts a truly AI-First approach where AI recommends the Level1 Plan based on the given pathway intention. Humans verify andmoderate these AI-generated plans through interactive dialoguewith AI, continuing this process through Level 2 (subtasks) andsubsequent hierarchy levels. At the task execution level, AIimplements the tasks while humans maintain oversight throughverification and validation of outcomes. This flexible approachensures the methodology is adaptable and can evolve alongside Amazon Web Services 3AI capabilities while maintaining human control over criticaldecisions.III. CORE FRAMEWORKThis section outlines the core framework of AI-DLC, detailing itsphases, roles, workflows, and key artifacts.1. ARTEFACTSAn Intent in AI-DLC is a high-level statement ofpurpose that encapsulates what needs to be achieved, whether abusiness goal, a feature, or a technical outcome (ex. performancescaling). It serves as the starting point for AI-drivendecomposition into actionable tasks, aligning human objectiveswith AI-generated plans.A Unit represents a cohesive, self-contained workelement derived from an Intent, specifically designed to delivermeasurable value. For instance, an Intent to implement abusiness idea may be decomposed into Units representingindependent functional blocks, analogous to Subdomains in DDDor Epics in Scrum. Each Unit encompasses a set of tasks (userstories, in this case, that articulate its functional scope) In thecontext of AI-DLC, the process of decomposing Intents into Unitsis driven by AI, with developers and/or Product Owners validatingand refining the resulting Units to ensure alignment with businessand technical objectives. The units are loosely coupled, enablingautonomous development and independent deploymentdownstream.A Bolt is the smallest iteration in AI-DLC, designed for therapid implementation of a Unit or a set of tasks within a Unit.Bolts (analogous to Sprints in Scrum) emphasize intense focusand high-velocity delivery, with build-validation cycles measuredin hours or days rather than weeks. Each Bolt encapsulates a well-defined scope of work (e.g., a collection of user stories within aUnit), enabling incremental progress while maintainingalignment with the overall objectives of the Unit it supports. AUnit can be executed through one or more Bolts, which may runin parallel or sequentially. AI will plan the Bolts with developers /Product Owners validating it.The Domain Design artefact models the core businesslogic of a Unit, independently of the infrastructure components.In the first version of AI-DLC, AI uses domain-driven designprinciples to create the strategic and tactical modelling elementsincluding aggregates, value objects, entities, domain events,repositories and factories. The Logical Design translates DomainDesigns by extending them for meeting the non-functionalrequirements using the right choice of architectural designpatterns (ex. CQRS, Circuit Breakers etc.). AI creates theArchitecture Decision Records (ADRs) for validation by theDevelopers. With the Logical Design specification, AI willgenerate the Code and Unit Tests, ensuring adherence to well-architected principles by selecting appropriate AWS services andconstructs. At this stage, the AI agent will conduct the unittesting, analyse the results and provide recommendations onfixes to the Developer.The Deployment Units are the operational artifactsencompassing the packaged executable code (ex. containerimages for Kubernetes environments, serverless functions suchas AWS Lambda), configurations (ex. Helm Charts), andinfrastructure components (ex. Terraform or CFN stacks) that aretested for functional acceptance, security, NFRs and other risks.AI generates all associated tests, including functional tests, staticand dynamic security tests, and load testing scenarios. After thehuman validation and adjustments on the test scenarios andcases, the AI agent executes the test suits, analyses the resultsand correlate failure points with code changes, configurations orother dependencies. Thus, these units are rigorously tested forfunctional acceptance, security compliance, adherence to non-functional requirements (NFRs), and mitigation of operationalrisks, guaranteeing their readiness for seamless deployment.2. PHASES & RITUALSThe Inception Phase focuses on capturing Intents andtranslating them into Units for development. This phase uses the“Mob Elaboration”, a collaborative requirements elaboration anddecomposition ritual. This happens in a single room with a sharedscreen led by a facilitator. During Mob Elaboration, AI plays a Amazon Web Services 4central role in proposing an initial breakdown of the Intent intoUser Stories, Acceptance Criteria and Units, leveraging domainknowledge, and the principles of loose coupling and highcohesion for rapid parallel execution downstream. The ProductOwner, Developers, QA and other relevant stakeholders (themob) collaboratively review and refine these AI-generatedartefacts by adjusting under-engineered or over-engineeredparts and aligning them with real-world constraints. The outputsof this phase include well defined Units and their respectivecomponents containing a) PRFAQ, b) User Stories, c) Non-Functional Requirement (NFR) definitions, d) Description of Risks(matching with organization’s Risk Register, if present), e)Measurement Criteria that traces to the business intent and thef) Suggested Bolts using which the Units can be constructed. MobElaboration condenses weeks or even months of sequential workinto a few hours, while achieving deep alignment both within themob and between the mob and the AI.The Construction Phase The encompasses theiterative execution of tasks, transforming the Units definedduring the Inception Phase into tested, operations-readyDeployment Units. This phase progresses through DomainDesign, where AI models the business logic independently oftechnical considerations, to Logical Design, where non-functionalrequirements and appropriate cloud design patterns are applied.AI generates detailed code from Logical Designs by mappingcomponents to appropriate AWS services while adhering to well-architected principles. The phase concludes with automatedtesting to ensure functionality, security, and operationalreadiness. Developers focus on validating AI-generated outputsat each step and making critical decisions, ensuring quality andadaptability in each iteration. In the brown-field (existingapplication) scenarios, the construction phase involves firstelevating the codes into a semantic rich modelling representationso that the context to AI becomes concise and accurate. Thesuggested modelling representations are static models (just thedomain components, responsibilities and their relationships) anddynamic models (how the components interact to realize thesignificant use cases)AI plays a pivotal role throughout this phase, recommendingtasks and providing options (design patterns, User Experience,Tests etc) at each task. AI-DLC recommends that this be done withall teams collocated in a single room, similar to Mob Elaboration.The teams exchange the integration specifications (from domainmodel stage), make decisions and deliver their bolts. AI-DLC callsthis the mob-construction ritual.The Operations Phase in AI-DLC centers on thedeployment, observability, and maintenance of systems,leveraging AI for operational efficiency. AI actively analyzestelemetry data, including metrics, logs, and traces, to detectpatterns, identify anomalies, and predict potential SLA violations,enabling proactive issue resolution. Additionally, AI integrateswith predefined incident runbooks, proposing actionablerecommendations such as resource scaling, performance tuning,or fault isolation and execute the resolutions when approved bythe Developers. Developers serve as validators, ensuring AI-generated insights and proposed actions align with SLAs andcompliance requirements.3. THE WORKFLOWGiven a business intent (ex. Green-Field development, Brown-Field enhancement, modernization, or defect fixing), AI-DLCbegins by prompting AI to generate a Level 1 Plan that outlinesthe workflow to implement the intent. This plan serves as aninitial proposal, which is then transparently reviewed, validated,and refined by humans to ensure alignment with business goalsand engineering constraints. At the heart of AI-DLC is theprinciple of applying human oversight to progressively enrich theartefacts of each step, transforming them into semantically richcontext for the next. Each step serves as a strategic decision pointwhere human oversight functions like a loss function - catchingand correcting errors early before they snowball downstream.This repeats recursively. Each step in the Level 1 Plan is furtherdecomposed into finer-grained, executable sub-tasks by AI, againunder human oversight to ensure accuracy and contextualappropriateness.All artefacts generated (intents, user stories, domain models, ortest plans) are persisted and serve as a “context memory” thatthe AI references across the lifecycle. Like traditional SDLCmethods, AI-DLC is inherently iterative, allowing for continuousrefinement and adaptation. Additionally, all artefacts are linked,allowing for backward and forward traceability (ex. connectingdomain model elements to specific user stories) ensuring that theAI retrieves the correct and most relevant context at every stage.Throughout the process, AI performs the strategic planning, taskdecomposition, generation etc. and humans provide theoversight and validation.IV. AI-DLC IN ACTION: Green-Field Development Amazon Web Services 5We will examine the scenario in which the Product Ownercommences the process by articulating a high-level intent, suchas "Develop a recommendation engine for cross-selling products."AI Recognizes this as an intent to build a new application andproduces the Level 1 plan like the Workflow steps in the abovesection. The team validates, verifies and adds/fixes the stages inthe level 1 plan. With the finalized Level 1 Plan, AI progresses tothe Inception Phase. Refer to the prompts in Appendix A as a wayto interact and provide oversights to AI.1. INCEPTION PHASEThe following bullet points outline the key interactions in theMob Elaboration ritual.a. AI asks clarifying questions (e.g., "Who are the primary users?What key business outcomes should this achieve?"), ensuringa comprehensive understanding of the goal and minimize theambiguity in the original intentionb. AI elaborates the clarified intention into user stories, non-functional requirements (NFRs), and risk descriptions. Theteam validates these artefacts and provides oversight, andthe corrections needed to AI.c. AI composes the highly cohesive stories into Units, e.g., "UserData Collection," "Recommendation Algorithm Selection,"and "API Integration."d. The Product Owner validates these outputs, adjustingwherever needed to refine the Units. Example: The ProductOwner notices that User Data Collection lacks privacycompliance details and adjusts the requirements to includeGDPR-specific considerations.e. AI generates a PRFAQ for the module (optional),summarizing the business intent, functionality, and expectedbenefits.f. Developers and Product Owners validate the PRFAQ andassociated risks, ensuring alignment with the overallobjectives.2. CONSTRUCTION PHASEThe following bullet points outline the key activities involved inthis phase, focusing on the Mob Programming and Mob Testingrituals.a. The Developer establishes the session with AI. AI prompts thedeveloper to begin with the Unit assigned to them.b. AI models the core business logic for the assigned Unit usingDomain-Driven Design principles. Example: For the"Recommendation Algorithm" Unit, AI identifies relevantentities like Product, Customer, and Purchase History andtheir relationships.c. Developers review and validate the domain models, refiningbusiness logic and ensuring alignment with real-worldscenarios (e.g., how to handle missing purchase history fornew customers)d. AI translates the domain models into logical designs, applyingNFRs like scalability and fault tolerance. Example: AIrecommends architectural patterns (e.g., event-drivendesign) and technologies (e.g., AWS Lambda for serverlesscomputation).e. Developers evaluate AI’s recommendations, approve trade-offs, and suggest additional considerations if needed. (e.g.,Accepts Lambda for scalability but overrides the storage toDynamoDB for faster query performance)f. AI generates executable code for each Unit, mapping logicalcomponents to specific AWS services.g. It also auto-generates functional, security, and performancetests (e.g., For the "Recommendation Algorithm" Unit, AIcreates code to implement collaborative filtering andintegrates it with a DynamoDB data source) Amazon Web Services 6h. Developers review the generated code and testscenarios/cases, making adjustments where necessary toensure quality and compliance.Testing and Validation:a. AI Executes all tests (functional, security, and performance),analyzes results, and highlights issues.b. Proposes fixes for failed tests, e.g., optimizing query logic forbetter performance.c. Developers validate AI’s findings, approve fixes, and reruntests as needed.3. OPERATIONS PHASEDeployment:a. AI packages the module into Deployment Units (e.g.,container images, serverless functions).b. Developers approve the deployment configuration andinitiate rollout to staging and production environments.Observability and Monitoring:a. AI analyzes metrics, logs, and traces to identify anomaliesand predict potential SLA violations. Example: AI detects alatency spike during peak usage and proposes scaling therecommendation engine to handle increased traffic.b. AI integrates with playbooks to suggest actions foroperational issues. If API response times degrade, AIrecommends increasing DynamoDB throughput orrebalancing API Gateway traffic.c. Developers validate AI’s recommendations, approvemitigations, and monitor resolution outcomes.V. AI-DLC IN ACTION: Brown-Field DevelopmentBrown-field refers to making changes to an existing system interms of either adding new features, optimizing it for non-functional requirements or fixing technical debts includingrefactoring and fixing defects. In this context, we will examine ascenario where the product manager needs to add a new featureto an existing application.1. INCEPTION PHASEThe inception phase activities in the brown-filed are same as thatof the green-field2. CONSTRUCTTION PHASEa. AI elevates the codes into a higher-level modellingrepresentation. The models comprise of static models(components, descriptions, responsibilities andrelationships) and dynamic models (how the componentsinteract to realize the most significant use cases)b. Developers collaborate with product managers to review,validate and correct the static and dynamic models that arereverse engineered by AI.c. With these extra steps, the rest of the construction phase issimilar to that of the green-field scenario.3. OPERATIONS PHASEThe operations phase activities in the brown-filed are same asthat of the green-fieldVI. ADOPTING AI-DLCAI-DLC does not deviate much from the existing Agile methodsand it is designed with easier adoption as a key outcome. Stillorganizations that are practicing the traditional methods forlonger and those who are in the process of inventing their ownvariation of AI-Native methods need specific strategies foradopting AI-DLC. We believe the following 2 approaches willmake this easier.a. Learning by Practicing – AI-DLC is actually a set of rituals (MobElaboration, Mob Construction etc.) that can be practiced asa group. Instead of learning the method via documentationand traditional trainings, we will get the practitioners topractice the rituals with the AI-DLC guides in multiple realworld scenarios that are currently being solved by thepractitioners. The AWS Solution Architects have created afield offering called AI-DLC Unicorn Gym that packages thisapproach for hyper-scaling adoption in large organizations.b. By embedding AI-DLC in the new Developer ExperienceTooling – our customers are building their own orchestrationtools that cut across SDLC, providing a unified experience fortheir developers. (ex. FlowSource from Cognizant, CodeSpellby Aspire, AIForce by HCL etc.) By embedding AI-DLC in thesetools, the developers in large organizations will seamlesslypractice AI-DLC without any need for significant adoptiondrives. Amazon Web Services 7APPENDIX AThe following prompts can be used to interact with AI forpracticing AI-DLC.##Setup PromptWe will work on building an application today. For every front endand backend component we will create a project folder. Alldocuments will reside in the aidlc-docs folder. Throughout oursession I'll ask you to plan your work ahead and create an md filefor the plan. You may work only after I approve said plan. Theseplans will always be stored in aidlc-docs/plans folder. You willcreate many types of documents in the md format. Requirement,features changes documents will reside in aidlc-docs/requirements folder. User stories must be stored in theaidlc-docs/story-artifacts folder. Architecture and Designdocuments must be stored in the aidlc-docs/design-artifactsfolder. All prompts in order must be stored in the aidlc-docs/prompts.md file. Confirm your understanding of thisprompt. Create the necessary folders and files for storage, if theydo not exist already.##Inception## User storiesYour Role: You are an expert product manager and are taskedwith creating well defined user stories that becomes the contractfor developing the system as mentioned in the Task sectionbelow. Plan for the work ahead and write your steps in an md file(user_stories_plan.md) with checkboxes for each step in the plan.If any step needs my clarification, add a note in the step to getmy confirmation. Do not make critical decisions on your own.Upon completing the plan, ask for my review and approval. Aftermy approval, you can go ahead to execute the same plan one stepat a time. Once you finish each step, mark the checkboxes as donein the plan.Your Task: Build user stories for the high-level requirement asdescribed here << describe product descrition>><<<After reviewing and changing the plan>>>>Yes, I like your plan as in the <<md file>>. Now exactly follow thesame plan. Interact with me as specified in the plan. Once youfinish each step, mark the checkboxes in the plan.## Units------Your Role: You are an experienced software architect. Before youstart the task as mentioned below, please do the planning andwrite your steps in the units_plan.md file with checkboxes againsteach step in the plan. If any step needs my clarification, pleaseadd it to the step to interact with me and get my confirmation.Do not make critical decisions on your own. Once you producethe plan, ask for my review and approval. After my approval, youcan go ahead to execute the same plan one step at a time. Onceyou finish each step, mark the checkboxes as done in the plan.Your Task: Refer to the user stories mvp_user_stories.md file.Group the user stories into multiple units that can be builtindependently. Each unit contains highly cohesive user storiesthat can be built by a single team. The units are loosely coupledwith each other. For each unit, write their respective user storiesand acceptance criteria in individual md files in the design/ folder.<<<After reviewing and changing the plan>I approve. Proceed.###Construction## Domain (component) model creation------Your Role: You are an experienced software engineer. Before youstart the task as mentioned below, please do the planning andwrite your steps in an design/component_model.md file withcheckboxes against each step in the plan. If any step needs myclarification, please add it to the step to interact with me and getmy confirmation. Do not make critical decisions on your own.Once you produce the plan, ask for my review and approval. Aftermy approval, you can go ahead to execute the same plan one stepat a time. Once you finish each step, mark the checkboxes as donein the plan.Your Task: Refer to the user stories in thedesign/seo_optimization_unit.md file. Design the componentmodel to implement all the user stories. This model shall containall the components, the attributes, the behaviours and how thecomponents interact to implement the user stories. Do notgenerate any codes yet. Write the component model into aseparate md file in the /design folder.<<<After reviewing and changing the plan>>>>I approve the plan. Proceed. After completing each step, mark thecheckbox in your plan file.##: Code Generation------Your Role: You are an experienced software engineer. Before youstart the task as mentioned below, please do the planning andwrite your steps in an md file with checkboxes against each stepin the plan. If any step needs my clarification, please add it to thestep to interact with me and get my confirmation. Do not makecritical decisions on your own. Once you produce the plan, ask formy review and approval. After my approval, you can go ahead to Amazon Web Services 8execute the same plan one step at a time. Once you finish eachstep, mark the checkboxes as done in the plan.Task: Refer to component design in thesearch_discovery/nlp_component.md file. Generate a verysimple and intuitive Python implementation for the NaturalLanguage Processing (NLP) Component that is in the design. Forthe processQuery(queryText) method, use amazon bedrock APIsto extract the entities from the query text. Generate the classesin respective individual files but keep them in `vocabMapper`directory.Refer to the generated codes in vocabMapper directory. I wantthe EntityExtractor component to make a call to GenAI. Thecurrent implementation uses the local vocabulary_repository.Can you analyse and give me a plan on how I can leverage GenAIfor both the Entity Extraction and Intent Extraction.##: Architecture------Your Role: You are an experienced Cloud Architect. Before youstart the task as mentioned below, please do the planning andwrite your steps in a deployment_plan.md file with checkboxesagainst each step in the plan. If any step needs my clarification,please add it to the step to interact with me and get myconfirmation. Do not make critical decisions on your own. Onceyou produce the plan, ask for my review and approval. After myapproval, you can go ahead to execute the same plan one step ata time. Once you finish each step, mark the checkboxes as donein the plan.Task: Refer component design model:design/core_component_model.md, units in the UNITS/ folder,cloud architecture in the ARCHITECTURE/ folder, and backendcode in the BACKEND/ folder. Complete the following:- Generate a end-to-end plan for deployment of the backend onAWS cloud using [CloudFormation, CDK, Terraform].- Document all the pre-requisites for the deployment, if any.Once I approve the plan:- Follow the best practice of clean, simple, explainable coding.- All output code goes in the DEPLOYMENT/ folder.- Validate that the generated code works as intended, by creatinga validation plan, generate a validation report.- Review the validation report and fix all identified issues, updatethe validation report.##: Build IaC/Rest APIs------Your Role: You are an experienced software engineer. Before youstart the task as mentioned below, please do the planning andwrite your steps in an md file with checkboxes against each stepin the plan. If any step needs my clarification, please add it to thestep to interact with me and get my confirmation. Do not makecritical decisions on your own. Once you produce the plan, ask formy review and approval. After my approval, you can go ahead toexecute the same plan one step at a time. Once you finish eachstep, mark the checkboxes as done in the plan.Task: Refer to the services.py under the construction/<>/ folder.Create python flask apis for each of the service there.