At Azul Linhas Aéreas, I identified a manual classification bottleneck and independently put a Claude API fine-tuning pipeline into production. It processes about 600 records a month and reduced work that took a week to five minutes.
At Nubank, I led the strategy for monitoring production ML models, covering data quality, latency, availability, and performance drift. I also built modular Python frameworks, mentored three analysts, and used dashboards to turn technical findings into clear business recommendations.
