Ramiro Cerdá
Senior Full-Stack.
Performance, observability, and quality. AI agents in production — architecture, code, and deploy.
01 / WORK
- Mundial 2026 — World Cup Companion Full-stack companion for World Cup 2026 end to end: groups, local-timezone fixtures, knockout bracket, venue map, and per-country TV guides. Historical matchup data crossed from StatsBomb and OpenFootball.
- Soccer Analytics Agent — Football analytics agent Chat agent over ~49k international matches (1872–today) built with a hand-written LLM tool loop, no agent framework. Gemini reasons and calls tools; Postgres + pgvector is the single layer for data, memory, and observability. A learning-first replication of Oracle's workshop on an open, GCP-deployable stack.
- AI Agents Hub — AI agent lab Learning and reference monorepo for Google ADK and LangGraph: from small agents (PlanReAct, output_schema, MCP, code executor) to full-stack apps in production. Includes Customer Support Chat and Career Coach (Vertex AI Agent Engine + Cloud Run).
- Gastify — Finance tracking with a Telegram bot A personal finance tracker with two entry points: a PWA dashboard and a Telegram bot that parses natural-language messages ('I spent 5k at the store on debit') into structured transactions using AI.