Portfolio
A collection of my personal projects, from frontend and backend to machine learning, games, and MCP. I like building things that actually run.
An AI chatbot with RAG (Retrieval Augmented Generation). It embeds the question, finds similar documents in pgvector, and streams an answer token by token from a PyTorch model that uses those documents as context. The whole stack runs with Docker Compose.
Draw a digit on the canvas and get a live prediction from a CNN trained with PyTorch. A pretrained model (99.27% test accuracy) is included, so a few commands are all you need. The frontend is plain HTML/CSS/JS with no build step.
Tetris with a single-player mode and a VS-CPU mode. The client is built with Unity, authentication and the ranking go through Nakama, and the CPU's moves come from a hand-written Go server. CI runs the Go checks, builds the Docker image, and runs the dotnet tests.
A memory (concentration) card game built with Unity, played with fruit pictures. It has a 4×4 board, card flip animation, and a move counter. When you clear a round, it sends your score to the companion API server and shows the ranking. It also works on its own when the server is not running.
The API that receives scores from the memory game above. It offers JWT authentication, a ranking per number of pairs, and interactive API docs with Scalar, and it runs with Docker Compose.
A small mobile-ordering system: browse a menu, put dishes in a cart, and place an order. It pairs a DDD-structured Laravel API with a Next.js mobile UI, all running in Docker. The cart lives in Redis, orders are stored in MySQL, and the API is documented with Swagger UI.
A minimal Model Context Protocol (MCP) server, written in Go, that exposes GitHub operations as tools.
Register it with Claude Code, ask in plain language, and it calls the list_issues tool
and summarizes the issues for you.
