Yusra Batool

Full-stack AI journaling companion built solo in 24 hours. LLaMA 3.3 70B via Groq for empathetic conversational AI, MongoDB Atlas for persistent mood and entry storage. Custom animated UI with login system and mood analytics.

5-agent pipeline (analyzer, matcher, cover letter writer, email, scheduler) using Google ADK SequentialAgent. Deployed on Cloud Run with Firestore. Separate MCP server via FastMCP exposing Firestore tools over SSE transport.

Multi-agent medical information assistant built with Google ADK and Gemini 2.5 Flash. Real-time Wikipedia research via LangChain. Architecture includes greeter, symptom researcher, and medical advisor agents deployed on Cloud Run.

AI-powered pneumonia detection from chest X-rays. 7-model weighted ensemble (DenseNet-121, EfficientNet-B4, ViT-B/16, ResNet-50, InceptionV3, MobileNetV3, AttentionCNN) achieving 98.4% ensemble accuracy. Grad-CAM heatmap overlays, severity grading, bacterial vs viral subtype classification, and bilingual PDF reports in English and Urdu. Doctor auth, patient records, and MongoDB Atlas integration.

GUI-based adventure game implementing BFS, DFS, Dijkstra's, and other graph traversal algorithms. Real-time pathfinding visualization, multiple difficulty levels, dynamic maze generation, and animated gameplay. Demonstrated at AI and CS Expo 2024.

Role-based LMS enabling student-faculty interaction with full CRUD functionality and structured academic reporting. Secure data flows, normalized SQL schema, user role management (admin/student/faculty), and modular OOP architecture.