I work on AI agents: agentic and harness engineering, agent and LLM evaluation, tool use, and Claude Code workflows. Senior AI Research Engineer at Mercado Libre and Claude Community Ambassador in Bogotá. I came up through evaluation (Inspect AI, trajectory grading) and production ML (recommender systems, ranking, computer vision) before moving to building agents. The models here are mostly GGUF quants from my local-inference experiments.