I’m a postgraduate student at the University of Cambridge, where I study human-inspired AI. My thesis focuses on designing multi-agent orchestration under real-world conditions, such as varying expertise, costs, and availability. Additionally, I am pursuing research on multi-agent reinforcement learning for political alignment. In the future, I aim to explore how such technologies can support ecological conservation through intelligent remote sensing and monitoring.
Previously, I worked as a Software Engineer at Stellar Cyber in San Jose, California, where I developed AI-driven interfaces for threat hunting [Stellar Cyber] [AIM Research, 2024] and human-augmented autonomous cybersecurity operations powered by agentic AI [Stellar Cyber] [Business Wire, 2025]. I'm also the creator of GONEXT, a generative AI platform that delivers personalized analytics for League of Legends players.
In my free time, I enjoy being outdoors in nature, going to music festivals, and playing racquet sports.
GONEXT is a League of Legends analytics platform built on a multi-agent LLM architecture. It provides transparent reasoning via a thinking trail and MCP logs, calculating detailed aggregate statistics from match history. The system offers context-aware strategies and optimized item builds based on live game states, while supporting dynamic conversational inquiries about anything game related, patches, players, and tournaments.
This open-source Model Context Protocol (MCP) server empowers LLMs with comprehensive access to League of Legends game data through the Riot Games API. It features over 35 tools and resources for retrieving player statistics, match history, champion information, tournament data, and real-time game monitoring, supporting both stdio and SSE transports for seamless client integration. It is shipped as a Python package and a Docker container. Additionally, a demonstrative MCP client chatbot, leveraging a ReACT agent, is also provided to showcase the server's extensive capabilities and practical usefulness.
docker pull kostadindev/league-mcp
pip install league-mcp
Python package for converting diverse content into a search-engine-friendly knowledge base. It effortlessly ingests files (PDFs, DOCXs, spreadsheets), websites, and GitHub repositories, then leverages LLMs to generate a Markdown knowledge base. Ideal for creating structured and crawlable formats like llms.txt and llms-full.txt, supporting Retrieval-Augmented Generation (RAG) applications, or synthesizing unstructured data from multiple sources. Capable of summarizing an entire github repostory or website in two lines of code.
pip install knowledge-base-builder
