Antonij Simonovski

Led design and implementation of a RAG-based command execution system enabling natural language control over backend workflows; embedded structured entities into a gRPC API and used semantic similarity to resolve entity references with low-latency vector search; built an embedding synchronization mechanism across schema and payload changes for real-time consistency of semantic search results; integrated Hugging Face and PyTorch with LangChain to manage LLM integration via Model Context Protocol and coordinate interactions within the command execution pipeline; delivered a full-stack content-based task recommender system achieving a 98% hit rate from the top 3 suggestions; developed a generic natural language query interface to retrieve accessible system data via MCP layer; reduced OpenAI API costs by 66% by internalizing analysis and optimizing LLM usage patterns across services.

Led design and implementation of a RAG-based command execution system enabling natural language control over backend workflows. Integrated Hugging Face and PyTorch with LangChain to manage LLM integration via Model Context Protocol (MCP); built entity embedding/version synchronization for real-time semantic search; delivered a content-based task-assignee recommender with a 98% hit rate among top 3 suggestions; created a natural-language data retrieval interface; reduced OpenAI API costs by 66% through optimization.

Transitioned the website from a no-code platform to a custom codebase using Vue.js, increasing profit margins by 10%; implemented user authentication, payments, and data storage with Supabase (PostgreSQL, Auth, Storage) and Python.