SELECTED WORKSELECTED WORK

• ✓Deep zoom functionality for gigapixel imagery (GeoTIFF and PSB).
• ✓Automated GDAL tile pyramid generation cutting browser memory consumption to <45MB.
• ✓Persistent vector annotation system pinned across continuous zoom levels.
• ✓Secure user authentication and role-based administrative workflows.
• ✓Cloudflare R2 integration for scalable cloud tile storage and fast global CDN delivery.
• ✓PostgreSQL and PostGIS database for persistent spatial metadata indexing.

Independently conceptualized, architected, and built the complete product: designed the automated GDAL ingestion pipeline to produce Deep Zoom Image (DZI) tile pyramids, implemented the FastAPI asynchronous tile server, configured PostGIS for spatial metadata indexing, built the OpenSeadragon viewport renderer with persistent vector annotation layers, and packaged the entire platform with Docker Compose.

• ✓Multi-source paper search across arXiv and OpenAlex in one query flow.
• ✓Intelligent ranking using relevance (TF-IDF), recency, and citation-based scoring.
• ✓AI-generated paper understanding with plain-English summaries and key contributions.
• ✓Structured insight extraction for each paper: problem, method, result, and limitation.
• ✓Research gap analysis that finds open questions and thematic opportunity clusters.
• ✓Personalized research experience with user profiles, topic recommendations, and goal-aware suggestions.

Led technical architecture and development: designed the dual-source ingestion engine for arXiv and OpenAlex, wrote the multi-factor scoring formula balancing TF-IDF lexical relevance, citation velocity, and publication recency, implemented an in-memory TTL caching tier to respect external API rate limits, and integrated the Google Gemini API with strict prompt contracts for verbatim-supported synthesis.

• ✓Drag-and-drop layer nodes and connect them to design a model visually.
• ✓Converts the visual graph into executable TensorFlow/Keras model code.
• ✓Supports built-in datasets like MNIST, Fashion-MNIST, CIFAR-10, CIFAR-100, and custom dataset uploads.
• ✓Lets users configure training and view metrics such as loss and accuracy over epochs in real time.
• ✓Trained models can be deployed as production API endpoints with generated API keys.
• ✓Provides an in-browser UI to test deployed models by sending inference requests and viewing predictions.

Led the project from conception to working prototype: created the interactive node-graph canvas using React Flow, engineered the topological sorting graph compiler in FastAPI that converts node JSON into valid TensorFlow/Keras Python scripts, and built the WebSocket-backed training monitoring interface that streams live epoch loss curves.

• ✓Personalized learning path generation from diagnostic quizzes.
• ✓Interactive 3D user interface designed with Spline.
• ✓Dynamic quiz generation with real-time mastery tracking.
• ✓Progress tracking and learning analytics dashboard.
• ✓Futuristic design system and responsive mobile learning experience.
• ✓Backend built with Node.js, Flask, and SQLite3.

Designed the responsive web interface, built the interactive syllabus node graph view, managed client-side mastery state with React Context, and implemented diagnostic quiz checkpoint evaluations.

• ✓AI-powered malaria cell detection using Keras & deep CNNs with 94%+ accuracy.
• ✓Hospital patient management system.
• ✓Automated clinical PDF medical report generation in under 30 seconds.
• ✓Real-time analytics dashboard with probability scoring breakdown.
• ✓Hospital laboratory triage workflow integration.
• ✓Secure cloud storage with Supabase & FastAPI.

Engineered the end-to-end computer vision pipeline and clinical web interface: trained a convolutional neural network (CNN) on segmented erythrocyte cell images, built an asynchronous FastAPI inference service for real-time sample processing, implemented confidence probability scoring (parasitized vs uninfected), and designed an automated clinical PDF test report generation workflow.