Roman Matveev

Led the full development cycle: performed reverse engineering of existing solutions, designed the API from scratch, and implemented an automated trading strategy.

Result: The bot delivered 8–30 % daily profit on balance before the platform started detecting bots.

Plans: Simulate a user in a pool of accounts with reduced activity for each; discover and connect new marketplaces.

Built and launched a FastAPI web platform that consolidates data on 300+ language models from the OpenRouter API — names, context-window length, and current token pricing — and automatically augments them with their positions from the LMArena leaderboard. The result is presented in an interactive searchable table.

Result: The platform is in production: aggregates 300 + open-source models, helping developers quickly find and compare LLMs.

Plans: Automatic tracking of new leaderboard positions and a public API.

Developed a chat assistant for developers: an LLM core and FastAPI server on MCP that hot-loads plugins and manages them, supporting toolchains with context passing.

Result: Delivered a ready-to-use EnhancedMCP core with plugin hot-reload, a CLI client with smart prompter and cascading calls; basic plugin set: file system, multi-LLM, browser automation, terminal, and chain-of-thought history.

Plans: UI development; I/O optimisation; improved prompt engineering (dynamic templates, token optimisation); extended auto-executability (plan-and-execute cycles).