Products, experiments and open-source work developed beyond client and production engagements.
Designed and built an MVP real-time voice platform with FastAPI, LiveKit/WebRTC and PostgreSQL, separating application control from the media layer and experimenting with AI voice participants.
Built a microservice-based AI product that extracts YouTube transcripts, summarizes and answers questions with LLMs, and exposes the workflow through a Telegram bot with authentication and payments.
Built a Telegram-based AI video generation product that turns user images into short vertical videos through fal.ai, with asynchronous processing, user limits and payment logic.
Built a job discovery tool that retrieves vacancies through internal APIs and uses LLM-based filtering to return roles matching a target technical profile.
Built an OCR-to-LLM pipeline for processing photographed tasks, combining image validation and text recognition with local and API-based language models.