Senior AI engineer / game developer
RuiciroRivera
I build production LLM systems by day — voice agents, retrieval, evaluation pipelines running against real traffic. By night I ship my own AI products, and a game. Both taught me the same thing: the hard part was never capability.
01 Writing
Six projects, one person, one agent: building a command-center on Claude Code
How I built Mithra — a personal command-center on Claude Code primitives. Memory, hooks, a skill that classifies work by blast radius, and an MCP server over a real multi-repo workspace. Plus why the security decisions mattered more than the features.
02 Work
AI & agentic systemsGames
Mithra
A command-center built on Claude Code primitives: an MCP server exposing six read-only tools over a real multi-repo workspace, a local GUI with an embedded terminal, and a permission model that decides what an agent is allowed to touch.
aibridal
A B2C wedding-planning app with an LLM assistant that runs discovery, builds moodboards and shortlists vendors. Freemium paywall enforced server-side. Takes real money from real couples.
ibride
The SaaS it spun out of: planning software for professional wedding planners. Built with my wife, who is one — so the requirements arrive from a real user, daily, whether I want them or not.
Condexia
Condo-administration platform — admin back office plus a resident PWA. Payments, bulk imports, multi-tenant isolation. Built with a business partner.
Strings
A narrative game in Unreal Engine 5 about a puppeteer and the people on his strings. Built solo under my own studio — design, blockout, cinematics. The long game, and the reason I care about tools that let one person hold a lot at once.