Danil Uvarov

At VAULTI, I’m co-developing an AI agent designed to work within a larger application ecosystem. I work on model selection, local LLM inference, and agent behavior, using llama.cpp with Qwen.

I’ve implemented tool-calling orchestration so the LLM can select and execute tools, receive results, and continue generating. The agent’s tool layer supports operations such as search, editing, and database interactions.

I’ve designed the agent around separate routes, views, handlers, and tools, and developed a Dockerized service that communicates with the project backend. The system is developed and tested for local, offline inference rather than external AI APIs.

At FLAVOURU, I developed a web application using React and Vite, integrated REST APIs through Axios, and implemented authentication-related functionality using JWT. I also worked across frontend and backend integration.