Projects

Full-stack platform for the NTI educational institution. Two programme tracks — grant incubation and live practice placement — covering applications, onboarding, project tracking, and reporting. Handles 9 user roles, notifications, dashboards, and a public CMS.

MCP server that gives coding agents a GUI for asking you questions. Built with Go + Wails v3, lives in the system tray, and supports multi-select, free-form input, and desktop notifications.

Configurable Discord bot for managing self-assignable roles. Roles are defined in a YAML file, supports single- and multi-select categories, buttons with current state, and hot reloading the config without restarting the bot.

Discord onboarding bot with a fully configurable application flow. Admins post a welcome embed with a persistent Get Started button — users fill out a modal, and the application routes to admins for review (Accept/Deny/Open-ticket) or directly creates a private ticket channel. Fully configurable via TOML — no code changes needed.

MCP server for sending messages to Discord via webhooks. Supports text, embeds, polls, files, link buttons, and forum threads.

Small Windows CLI for controlling the keyboard backlight on ASUS ROG laptops. Supports brightness levels from 0 to 3 and static RGB colors via ASUS Aura HID, with a ready-to-use Windows binary.

CLI tool for controlling Windows windows — list, focus, minimize, close. Go rewrite of WindowManager using Cobra + Win32 API for compact binaries and JSON output.

Self-hosted infrastructure on a VPS — Docker, Caddy, Tailscale VPN, Jellyfin, Navidrome, Portainer, monitoring (Beszel, Uptime Kuma, OpenObserve), and fail2ban. Runs my websites, media server, and personal services.

Tiny TUI for force-killing Windows processes. Shows only real application windows, skips background and system processes — single portable binary.

Semester project for the "Frontend Technologies" course. SPA with a collection of productivity lifehacks. Lifehacks are stored in JSON. Supports filtering by category, viewing details, and saving favorites.

This website. React SPA with Tailwind, 4-language i18n, and build-time prerendering for SEO. Deployed via GitHub Actions — pushes to main trigger a build, prerender 5 pages, and rsync to a Caddy-served VPS. Zero manual steps.

Study project building a Data Warehouse using the Kimball methodology on an NFL Fantasy Football dataset in Snowflake. Implements a star schema with SCD Type 2 and analytical queries for player ranking and fantasy points.

WebView app for speech-to-text using Whisper. Visualizes audio, uses VAD, and lets you define instructions to polish the text.

Published Firefox extension that cleans Genius.com pages from annotations, popups, and highlights, leaving only lyrics. Works with scrolling and dynamic content loading.

Online generator of custom CSS styles for the Pear Desktop web app, letting users hide unwanted UI elements in YouTube Music such as mood chips, social buttons, menu items, and more.

Tool for bulk downloading emojis and stickers from public Discord servers. Supports both static and animated emojis, collects metadata, and organizes results into JSON and folders.

Unofficial FastAPI client for MyInstants. Lets you search sounds by keywords, fetch trends, recently added sounds, and details for specific audio. Works via web scraping.

Lightweight utility for managing Windows windows from the command line. Supports listing windows, checking whether they are running, minimizing, and activating the target window. Includes a compiled binary via Nuitka.

EmoteLab mod that adds a second export button and encodes scenes with ffmpeg on every CPU core — same files, roughly 5–6× faster. Only the encoding step is replaced; the game's own export settings keep working.

Windows tray icon that shows how many tasks are left in Super Productivity without switching windows — orange for today, blue for the total, green when nothing is left. Left-click opens the app, and the local REST API token is found automatically.

Windows voice assistant: a hotkey starts listening, AssemblyAI transcribes the speech, an OpenAI-compatible model answers through 43 tools (apps, windows, media, OBS, Steam, web search, Super Productivity tasks), and Gemini TTS speaks the reply. Persistent memory and 216 automated tests.