At Pandatron, I work on an AI-powered coaching platform for enterprise clients. I built an LLM-powered reporting pipeline that turns coaching session data into reviewed, publish-ready insights, delivering more than four monthly reports for four enterprise clients.
I also reduced the session chat metrics API's average response time by adding in-memory caching, and made chat non-blocking while LLM responses stream by queuing user input.
At UNIMED CEARA, I improved Vue application performance with a custom cache controlled by API max-age headers, reducing page response time from 100 to 70 ms. I also created end-to-end tests and an observability system linked to a team-wide Telegram channel.
Earlier, at NESTEC, I worked on React and Next applications, adapting architecture to reuse components and adding request caching. I also launched a secure payment system using credit card tokenization with the Mercado Pago gateway API.

