At Méliuz, I architected a modular AI tooling platform on the Anthropic Claude API, delivering reusable skills for meeting summaries, task extraction, and automated pull request reviews. I also designed its team-specific configuration and TypeScript CLI orchestrator.
I integrated the platform with Slack, Jira, and Notion, and led adoption of Model Context Protocol servers, including a custom design for EAN/SKU product matching. I also built n8n workflows for affiliate email classification and interactive Slack notifications.
My work at Méliuz has also included automating Operations processes, monitoring business KPIs, and building data pipelines for UX/UI, e-commerce, and finance teams. In my research, I developed a PyTorch Transformer model for appliance-level energy disaggregation and prepared Brazilian Portuguese speech data for Distil-Whisper fine-tuning.
