I am a 3rd-year student at KhAI University. Proficient in Python, Git, NumPy, Docker, and various other technologies. Continuously improving my skills and striving to create efficient code and project architectures.

• 🌱 Currently, I'm studying Machine Learning
• 📫 You can reach me at: werior36@gmail.com

Feel free to connect with me for collaboration opportunities, project discussions, or just to say hello! I'm always open to new ideas and challenges.

Let's build something amazing together!

• Ukrainian (native)
• Russian (native)
• English (B1+)

[October 2023] python-tg-bot-report Python, telegram api

• Developed a Telegram bot that forwards user messages to a moderation group for manual review.
• Included basic anti-spam protection via keyword bans and user blocking.
• Lightweight and functional; created as a base solution for custom bot workflows.

[2024] Side Project: Web service for Dungeons & Dragons gameplay
Python, Flask, FastAPI, PostgreSQL, WebSockets, Docker, Google api, JavaScript, Go, Redis, Git, Kubernetes, CLI, Rest API, Linux (Ubuntu), Jira

• Built a custom game board engine for Dungeons & Dragons with real-time token movement, image uploads, editing, and map management. All player actions were synchronized via WebSocket.
• Designed and implemented the rendering and interaction logic from scratch, without using third-party engines. Thanks to deep optimization, the client could handle 50,000+ objects without lag on a standard PC.
• Developed a hybrid storage system: PostgreSQL for persistent game data, Redis for active session caching with lazy loading and unloading depending on activity.
• Integrated analytics via Redis to track player actions and implemented a set of microservices — from image handling to campaign sharing and invitation-based access.
• Enabled seamless hot updates with Docker during live sessions. Paid special attention to security, including session isolation and API protection.
• The platform was used in real D&D campaigns, but development was paused after 5 months due to shifting priorities and lack of monetization. Despite that, it provided hands-on experience in building complex web services, managing real-time systems, and maintaining production-grade infrastructure.

[2025] Few-Shot Text Classifier on PyTorch Python, pandas, numpy, pytorch -Built a 250M-parameter text classification model in PyTorch, optimized for few-shot learning scenarios (1 sample per class). Leveraged a custom vectorizer and high-dimensional representations to reach >85% accuracy. Explored lightweight training regimes for large models with limited data.

[2025] Telegram bot HAISchedule Python, FastAPI, Redis, PostgreSQL, Docker, Javascript, aiogram, web parsing, Git, CLI, Grafana, Linux (Ubuntu), Jira A Telegram bot for student class schedules, serving over 120 users, with more than 40 daily active users during the academic period. It uses a PostgreSQL database and Redis for analytics. The bot is implemented in Python and includes SSL protection for the database, ensuring secure external access.

[2025] Started development of personal website cv.harkushyn.com HTML, CSS, JavaScript, Python, FastAPI ...