Projects

Owned the read path for a high-traffic list of medical exams that was timing out under load. Profiled, identified Oracle JOIN cost, mirrored the read model into MongoDB, and got p95 from 7–13s down to 500–800ms. Also built a Lambda table view for fast COVID-19 results.

What I'd do differently nowToday I'd document the trade-offs with a one-pager so the rest of the org could replicate the pattern, instead of letting the win stay tribal knowledge.

Part of the first team worldwide to deploy React.js in production (2015), on a checkout handling 2M+ daily accesses. Also worked on a Django social-commerce platform that raised R$30M in 2017.

What I'd do differently nowInvest earlier in DX primitives — a shared component library would have saved everyone a year.

High-density data filtering on a complex dashboard, page-load improvements, and feature-flagged rollouts on high-traffic Inditex surfaces.

What I'd do differently nowPair every perf win with a CI guardrail so it doesn't regress the next sprint.

A shared dropdown component, a mock interceptor for the test suite, keyboard avoidance, and pipeline notifications. Individually small; compounded into noticeable team velocity.

What I'd do differently nowWrite down the patterns so newcomers can extend them without asking.

Integrated AI into product search; implemented a CDN strategy with Cloudflare + Azure Blob/Event Grid/Functions; shipped feature flags and CodePush for the mobile app.

What I'd do differently nowWrap the AI ranking in an explicit evaluation harness from day one.