Automation, data and AI agents: systems where you can trust the number.

What I build

• Business rules in Postgres/Supabase. The AI reads and ranks; the database calculates.
• n8n in production: scheduled pipelines with monitoring, rate-limit and timeout handling, heavy loads moved out of n8n.
• AI agents with tools, RAG with pgvector, structured outputs with evidence, cost measured per token.

Recent work

• Supplier price-list matcher for a supermarket in Argentina (delivered Oct 2026): spreadsheets, PDFs and photos matched to the store catalogue, with zero false positives as the acceptance criterion. 4,026 of 4,929 products identified across 74 suppliers. Case study and a public, self-testing demo (131 tests): https://github.com/rodrigoteixeiraferreira/comparador-listas-fornecedores
• Hotel booking agent in n8n (Spanish): RAG with pgvector, atomic booking in a Postgres function, the AI never confirms a payment. US$ 0.032 per full conversation: https://github.com/rodrigoteixeiraferreira/agente-hoteleiro-ia
• Telecom pre-sales agent in n8n: an orchestrator and 3 sub-agents sharing one vector base: https://github.com/rodrigoteixeiraferreira/agente-prevendas-telecom-n8n

Stack: n8n · PostgreSQL/Supabase · Python · SQL · TypeScript/Next.js · Anthropic API and Claude Code (MCP, subagents) · Gemini · Docker · Fly.io · Vercel

Based in Argentina · Portuguese, Spanish and English (written) · Open to projects · https://www.linkedin.com/in/rodrigo-teixeira-ferreira-2b00a31b6/