01 / 09FeaturedKiranaWalaAI-Powered Hyperlocal Grocery MarketplaceBuilt a MERN-stack grocery marketplace featuring 80+ products, incorporating feedback from five local store owners to validate real-world requirements.Secured payment processing with Razorpay signature verification and HMAC-SHA256 webhook validation to authenticate payment events and prevent duplicate processing.Integrated an AI-powered basket builder that converts natural-language grocery lists into structured carts, alongside demand prediction for inventory planning.ReactNode.jsExpressMongoDBRazorpayJestLive DemoGitHubCase Study →
02 / 09FeaturedBeyond the Rumour MillEvidence-Driven Information Analysis PlatformBuilt a full-stack information platform using Next.js, TypeScript, and PostgreSQL to help users examine claims and organize supporting evidence.Implemented a relational database layer with Drizzle ORM, database migrations, and persistent storage to support maintainable server-side application workflows.Integrated Vitest and GitHub Actions CI to automate testing, TypeScript checks, ESLint validation, and production-build verification.Next.jsReactTypeScriptPostgreSQLDrizzle ORMVitestDockerGitHubCase Study →
03 / 09FeaturedAskPDFAI-Powered Conversational Document AssistantBuilt a PDF question-answering application using retrieval-augmented generation (RAG), supporting documents up to 20 MB with page-level citations linked directly to the in-app PDF viewer.Engineered asynchronous document ingestion with Redis-backed workers, HTTP 202 responses, job tracking, and stale-job cleanup to separate long-running processing from API requests.Tuned semantic retrieval with a 0.7 cosine-similarity threshold to filter low-relevance passages before generation and improve the grounding of document-based answers.ReactTypeScriptNode.jsExpressMongoDBRedisGemini APILive DemoGitHubCase Study →
04 / 09FeaturedVoxShieldAI Voice Deepfake Detection & Audio ForensicsDeveloped a full-stack audio-forensics platform supporting uploaded-audio analysis through REST APIs and real-time microphone inference over WebSockets.Designed a multi-scale convolutional neural network with parallel branches using kernel sizes of 5, 11, and 21 to capture audio patterns at different temporal scales from 16 kHz raw-waveform input.Reported 96.42% accuracy and 0.9892 AUC across 4,850 samples, with a forensic dashboard featuring waveform visualization, 64-band log-Mel spectrograms, and five acoustic descriptors.PyTorchFastAPIReactTypeScriptWebSocketsLive DemoGitHubCase Study →
05 / 09Research PaperJalDrishti 2030AI-IoT Digital Twin for Predictive Water-Stress PlanningAuthored a research paper on applying AI-driven digital twins to urban water-stress prediction and intervention planning for Bengaluru.Investigated predictive analytics, IoT-based water monitoring, and digital-twin modeling for sustainable urban resource management.Reached the international semi-finalist stage of the SMART CITY 2030 student and young researcher competition.AI/MLIoTDigital TwinsUrban SustainabilityLive DemoGitHubCase Study →
06 / 092025MyStringCustom String Implementation in C++Implemented a custom C++ string class to manage character sequences and provide string-handling functionality beyond a basic application-level wrapper.Extended the class with dynamic memory management, constructors, destructor, and overloaded operators for string manipulation and stream I/O.Built to strengthen object-oriented programming fundamentals and low-level resource management.C++OOPDynamic Memory ManagementGitHubCase Study →
07 / 092025FixMyWayMobile Civic Issue Reporting ApplicationBuilt a React Native application with Expo and Firebase to let citizens report local infrastructure problems through a mobile interface.Implemented photo uploads, GPS geotagging, and status tracking for community-reported infrastructure issues.Won 2nd place at the TRINETRA Hackathon 2025, in an event with 600+ student registrations.React NativeExpoFirebaseLive DemoGitHubCase Study →
08 / 092025Crop AnalysisMulti-Temporal NDVI Crop Cycle AnalysisBuilt a web platform for extracting and analysing crop-cycle parameters from multi-temporal NDVI data.Connected frontend workflows with backend processing and data services for agricultural analysis.Presented crop-cycle information through a practical interface focused on interpretable results.ReactNode.jsExpress.jsMongoDBNDVIGitHubCase Study →
09 / 092026Portfolio v3Interactive Engineering PortfolioPersonal developer portfolio focused on immersive interaction and motion.Built with a modern Next.js and TypeScript stack.Customized the visual system and content around Jyatin's development profile.Next.jsTypeScriptTailwind CSSFramer MotionGSAPGitHubCase Study →