At Terminal, I evaluate AI-generated commit messages and pull request descriptions against code diffs, problem statements, and test results. I fact-check model claims and provide evidence-based preference judgments that feed into model training.
At Mercado Libre — Mercado Pago, I designed and maintained Go and Java microservices on AWS that processed 8–10 million Visa transactions daily across four countries. I also built a BigQueue retry system that recovered around 200,000 failed messages per day with zero message loss; at BRQ — Client: Itaú Unibanco, I built serverless Go and Python services for a bank-wide AWS migration platform.
