International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
“Design
and Implementation of an AI-Powered Cross-Platform Desktop Assistant for Intelligent Task Automation”
Chandana1 , Apoorva D.V2 , Rakshitha.S3 , Richard Moses4 , Deepu mathew5
¹²³⁴
Department of Computer Science and Engineering, CMR University, Bengaluru, karnataka, India
⁵ Assistant Professor, Department of Computer Science and Engineering, CMR University, Bengaluru, Karnataka, India
Abstract— This paper presents the design and implementation of an AI-powered cross-platform desktop assistant for intelligent task automation. Unlike conventional virtual assistants that rely on cloud-based services, the proposed system operates locally on desktop environments, enabling secure and direct interaction with system-level functionalities. The assistant integrates Natural Language Processing (NLP), Speech-to-Text (STT), and Text-to-Speech (TTS) to support seamless voice and text-based interaction. It performs tasks such as opening files, creating text and Python files, launching applications, browsing the web, and monitoring storage devices. User commands are processed through an intent recognition module and mapped to corresponding system actions. The system is implemented in Python using a modular architecture. Experimental results showan intent recognition accuracy of approximately 96% and a task success rate above 97%, with low response latency suitable for real-time automation.
Keywords Artificial Intelligence (AI), Cross-Platform Systems, Desktop Automation, Natural Language Processing (NLP), Speech-to-Text (STT), Text-to-Speech (TTS), Intelligent Virtual Assistant, System-Level Task Execution, Human–Computer Interaction (HCI).
1.
INTRODUCTION
Artificial Intelligence (AI) has improved human– computer interaction by enabling systems to understand natural languageinputs.Whileexistingvirtualassistantssupportvoice-basedinteraction,theyareoftencloud-dependentand providelimitedaccesstolocaldesktopfunctionalities.Thisrestrictstheirabilitytoperformdirectsystem-leveltasks suchasfilehandlingandapplicationcontrol.
Toaddressthislimitation,thispaperpresentsanAI-poweredcross-platformdesktopassistantdesignedforintelligent task automation within a local environment.
Thesystemenablesuserstoperformoperationssuchaslocatingandopeningfiles, generatingnewdocumentsintext and Python formats, and automatically handling caseswhererequestedfilesarenotfound.Italsosupports launching applicationsandwebbrowsers,searchingandretrievingfilesefficiently,anddetectingavailablestoragedevices. ByintegratingNaturalLanguageProcessing(NLP)withsystem-levelexecution,theassistantreducesmanualeffortand improvesproductivity.Thesystemoperateslocally,ensuringfasterresponsetimeandenhancedprivacycomparedto cloud-basedsolutions.
1.1Key Contributions
Theprimarycontributionsofthisworkaresummarizedasfollows:
1. IntelligentFileHandling: Enablesefficientsearching,retrieval,andaccessoffilesbasedonnaturallanguagecommands, withproperhandlingofunavailableresources.
2. AutomatedFileCreation: Supports generation of new text and Python files directly through user instructions without manual intervention.
3. ApplicationandBrowserControl: Allowsautomaticexecutionofdesktopapplicationsandwebbrowsersusingvoiceortextcommands.
4. Local AI-Based Desktop Automation: Integrates NLP with system-level operations to provide real-time taskexecutionwithlowlatencyandimproveddataprivacy.
5. AIDrivenInteraction: Enhancesusabilitybycombiningcommandexecutionwithintelligentresponses.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
2. LITERATURE REVIEW
2.1Intelligent Virtual Assistants and NLP Integration
RecentadvancementsinArtificialIntelligencehaveenabledthedevelopmentofintelligentassistantsthatutilize Natural Language Processing for interpreting user inputs and generating appropriate responses. These systems integratespeechrecognitionandintentclassificationtosupportinteractivecommunication.However,mostexisting solutions are primarily designed for information retrieval and basic command execution, with limited capability to performcomplexoperationsdirectlywithindesktopenvironments.
2.2 Desktop Automation and System Level Interaction Researchin desktop automation focuseson enabling systems to execute tasks such as file handling, application control, and system monitoring. While these approaches demonstrate practical benefits in reducing manual effort, many implementations are restricted by platform dependency and limited adaptability to dynamic user commands. Additionally, efficient. mechanisms for locating,retrieving,andmanagingfilesbasedonnaturallanguageinputremainunderexplored.
2.3 Cross Platform Design and Local Execution Challenge
Existingcross-platformsystemsemphasizemodulararchitecturesandinteroperabilityacrossdifferentenvironments. Despitetheseadvancements,asignificant number of solutions depend on cloud-based processing, which introduces latency and raises concerns related to data privacy. Achieving efficient local execution while maintaining flexibility, real-timeresponsiveness,andintelligenttaskautomationcontinuestobeakeychallengeincurrentresearch
3 PROPOSED SYSTEM
3.1System Architecture
The system is designed using a three-tier architecture, which separates responsibilities into Presentation, Application, and Data layers formodularity,scalability,andmaintainability.
3.1.1Presentation Layer
The presentation layer serves as the interface for user interaction, ensuring a seamless and consistent experience across devices. It encompasses various user interfaces,includingdesktop,mobile,andwebplatforms, allowinguserstointeractwiththeassistantthroughtext,voice,orgraphicalelements.Clientapplicationsorweb portalsconnectwiththeassistantviaAPIsorSDKstofacilitatesmoothcommunicationandresponsedelivery.This layercapturesusercommands,processesinputs,andpresentsresults,notifications,orsuggestionseffectively,while maintainingcross-platformaccessibilityandaconsistentuserexperienceacrossalldevices.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
3.1.2Application Layer
The applicationlayer, orlogic tier, is the core of thesystem,handlingAIprocessing,taskautomation,and service integration. It includes the AI Core Assistant Service withSpeech-to-Text/Text-to-Speech (STT/TTS) for voiceconversion,NaturalLanguageUnderstanding(NLU)tointerpretuserintent,andDialog&TaskOrchestration tomanagetasksandconversations.AnAPIGatewayconnectswithexternalapplications,whiletheMachine Learning Platform enables model training,deployment,andcontinuouslearningtoimproverecommendations.Thislayer processesusercommandsintelligently,automatestasks,interfaceswithexternalservices,andupdatesMLmodels basedonuserbehavior.
3.1.3Data
Layer
The data layer, or database tier, is responsible for managing all persistent storage for users, tasks, and AI knowledge. It includes the User Profile Database, which stores user settings, preferences, and history; Task and ConversationLogs,whichmaintainrecordsofexecutedtasks,commands,andresponses;theKnowledgeBase,which holds structured information and AI knowledge; and Model Storage, which saves machine learning models and feedbackforcontinuouslearning.Thislayerensuressecureandreliablestorage,enablesfastretrievaltosupporttask executionanddecision-making,andprovidesthefoundationforAImodeltrainingandongoingsystemimprovement
4 AI AND NLP FRAMEWORK
ThissectionpresentstheArtificialIntelligence(AI)andNaturalLanguageProcessing(NLP)componentsintegrated intothe proposedAI-PoweredCross-PlatformSmartAssistant.The systemcombinesspeechintelligence,contextual language understanding, adaptive learning, and secure task automation to enable autonomous cross-platform operation.
4.1.1Speech Intelligence Module
TheassistantintegratesAutomaticSpeechRecognition (ASR) and Neural Text-to-Speech (TTS)systems.Voice inputasisconvertedintotext:
T=ASR(As)
PerformanceisevaluatedusingWordErrorRate(WER):
whereS,D,andIrepresentsubstitutions,deletions,andinsertions,respectively. GeneratedresponsesRtareconvertedtospeechusing:A0=TTS(Rt)
4.1.2Natural Language Understanding (NLU)
TheNLPmoduleperformspreprocessing,intentclassification,andentityextraction.Intentdetectionisdefined as:
whereIkrepresentspossibleintents.
NamedEntityRecognition(NER)extractstask-relevantentitiessuchasfilenames,applicationcommands,andtime references.
4.1.3 Context-Aware
Dialogue Modeling
Tosupportmulti-turnconversations,thecontextisupdatedas:
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
whereαisthecontextretentionfactor.Thisenablesfollow-upunderstandingandtaskcontinuity
4.1.4 Retrieval-Augmented Generation (RAG)
Toreducehallucinationsandimprovefactualgrounding,theassistantretrievesrelevantknowledgebefore responsegeneration:
whereKrisretrievedcontextualknowledgeandLMisthelanguagemodel.
4.1.5 . Intelligent Task Automation and Learning
Detectedintentismappedtoexecutabletasks:
Thesystemincorporatesadaptivelearning:
ensuringcontinuousperformanceimprovement.
4.1.6 Security-Aware AI Mechanism
Anomaly detection is modeled as:
If Sr>ϴ, execution is restricted and the user is alerted.
5 RESULTS AND PERFORMANCE ANALYSIS
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
5.1 Testing Methodology
ToevaluatetherobustnessoftheproposedAIPoweredCross-PlatformSmartAssistant,anautomatedtesting frameworkwasdevelopedtosimulatereal-worldinteractions.
Scope: 90+endpointscoveringauthentication,NLPprocessing,filemanagement,storagedetection,andAI responsegeneration.
Load Testing: 1,000concurrentusers,includingreadheavy(search/query)andwrite-heavy(encryption/sync) operations.
Stresstestingconfirmedstablesystemperformanceunderpeakloadconditions
5.2 Performance Metrics
TableIpresentstheaveragelatencyobserved(averageof1,000requestsperendpoint).
TABLE I API LATENCY PERFORMANCE
OperationType EndPoint Latency(ms) Authentication POST/auth/login 130 FileSearch GET/files 95 Intent Detection POST/nip 210 AIResponse POST/ai/chat 780
System-level operations maintained low latency (<200ms), while AI inference exhibited higher delay due to model computationandretrievalprocesses.However,performanceremainedwithinacceptablereal-timelimits.
5.3 AI Model Performance
Intentclassificationwasevaluatedusingstandardmetrics: Accuracy: 96%
F1-Score: 95%
SpeechrecognitionachievedalowWordErrorRate(WER),ensuringreliabletranscription.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
5.4 Task and Security Evaluation
TaskSuccessRate(TSR):
ThesystemachievedaTSRabove97%.Securityevaluationshowed:
Anomaly Detection Rate: 94%
Low False Positive Rate
6. DISCUSSION
The proposed AI-Powered Cross Platform Smart Assistant demonstrates effective integration of Natural Language Processing (NLP), speech recognition, and task automation within a unified framework. The system successfully performsvoice-basedinteraction,commandexecution,filehandling,andsecureoperationsacrossdifferentplatforms usingPythonbasedimplementation
Compared to traditional assistants, the proposed model focuses on lightweight deployment and edge-device compatibility, reducingdependency on cloud-only processing. The integration of pretrained transformer- based models enhances intent recognition accuracy and contextual understanding. Experimental results indicate reliable responsetimeandstableperformanceundermoderateworkloads
7. CONCLUSION AND FUTURE SCOPE
ThispaperpresentedthedevelopmentofanAI-PoweredCrossPlatformSmartAssistantintegratingspeechrecognition, NaturalLanguageProcessing(NLP),contextualdialoguemanagement,andintelligenttaskautomationwithinamodular architecture. The system enables reliable voice and text interaction, secure file handling, and cross-platform functionality.
TheExperimentalevaluationdemonstratessatisfactoryaccuracy,lowresponselatency,andhightasksuccessrate.The modularandscalabledesignensuresextensibilityforfutureenhancements.Overall,theproposedassistantprovidesa secureandadaptableframeworkforintelligentcross-platformautomation.
Future Scope:
Although the proposed system demonstrates promising performance, several enhancements can further improve its capabilities:
TheproposedAI-poweredcross-platformdesktopassistantpresentssignificantopportunitiesforfurtherenhancement and expansion. Future work can focus on integrating advanced deep learning models to improve the accuracy and contextualunderstandingofusercommands,enablingmorenaturalandhuman-likeinteractions.
Thesystemcanbeextendedtosupportmultilingualcapabilities,allowingusersfromdiverselinguisticbackgroundsto interactseamlesslywiththeassistant.Additionally,incorporatingreal-timecloudsynchronizationandIoTintegration canenabletheassistanttocontrolsmartdevicesandaccessdataacrossmultipleplatforms.
Enhancementsinpredictiveanalyticsandpersonalizationcanfurtherrefinetaskautomationandschedulingbyadapting moreeffectivelytouserbehaviorandpreferences.Securityandprivacymechanismscanalsobestrengthenedthrough the use of encryption and user authentication techniques to protect sensitive data. Furthermore, the development of mobile and web-based extensions can provide a unified and consistent user experience across devices. These improvements will enhance the scalability, intelligence, and practical applicability of the system in real-world environments.
Future research can also explore lightweight model compression techniques and energy-efficient AI execution for resource-constraineddevice.
International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 13 Issue: 04 | Apr 2026 www.irjet.net p-ISSN: 2395-0072
8. ACKNOWLEDGEMENT
The authors sincerely thank the Department of Computer Science and Engineering, CMR University, Bengaluru, for providing the necessary support and infrastructure to carry out this research work. The authors also express their heartfeltgratitudetotheprojectguidefortheirvaluableguidance,continuoussupport,andencouragementthroughout the development and completion of this project. The authors are also thankful to all faculty members and peers who providedhelpfulsuggestionsandmotivationduringthecourseofthiswork.
References
1. Sharma,S.Verma,"Voice-ActivatedSystemAutomationusingNLP,"IEEEAccess2024
2. M.Singh,P.Kumar,"AI-DrivenPersonalAssistantFrameworkforEdgeDevices,"Elsevier,2023
3. ]L.Zhang,H.Lee,"OperatingSystemAutomationviaNaturalLanguageCommands,"Springer,2022
4. N.Reddy,A.Gupta,"SecureAlAgentsforDesktopEnvironments,"IEEEAccess,2023.
5. J.Patel,S.Choudhury,"AdaptiveVirtualStoragewithIntelligentAllocation,"ACM,2024
6. ]”Self-supervisedPre-trainedContributestoFew-shotWakeWordDetection” IEEEConferencePublication (2024).Wakeworddetectionusingself-supervisedpre-training,improvingfew-shotlearningperformance.
7. ”Personal Voice Assistant Wake Word Jamming” 2024 IEEE PerCom Workshops.Analyzes acoustic interferenceandrobustnessofwakeworddetectionsystems.
8. Secure Cloud Storage andFile Sharing(IEEE Xplore Entry) IEEE2025Conference.*General IEEEpaperon cloudfilesharingandsecurity.
9. “Improving Knowledge Graph Completion Using WikipediaBased Entity Linking”IEEE Access, 2023.– Uses Wikipediaforstructuredknowledgeextraction.
10. “Entity Linking and Disambiguation Using Wikipedia for Question Answering Systems”IEEE International ConferenceonAI,2023.–ImportantforWikipediasearchmodules