Melissa Merve Abdula

A 3D virtual fitness trainer that demonstrates exercises, tracks sessions, and supports real-time motion capture for form feedback. Built with Godot Engine, the assistant organizes routines by category (warm-up, arms, abs, legs, stretching), uses an interactive character for precise demonstrations, and includes timers, progress indicators, and smooth camera controls for optimal viewing angles. Blender was used for modeling/animation pipelines, while Python/OpenCV powered motion capture integration and preprocessing. Designed to be lightweight and responsive, the app delivers clear guidance, visual feedback, and consistent pacing — making solo workouts more structured and engaging.

A research‑driven build of a multilayer web application focused on application and database security. The work covers threat modeling, authentication and access control, secure configuration, input validation, and database hardening for PostgreSQL. It also includes containerized dev environment, reproducible builds, and deployment considerations.

An intelligent, document‑powered chatbot for Agency Software products. Ingests PDFs, DOC/DOCX, and HTML (with OCR via Tesseract/Azure) and uses Retrieval‑Augmented Generation backed by PostgreSQL+pgvector for semantic search. The FastAPI backend and React+TypeScript frontend deliver an interactive chat with conversation history and clickable source citations for every answer. Includes auth/security primitives, Redis for memory/caching, Alembic migrations, health checks, and Docker Compose for local/production deployment.