Selected projects
Client projects I took part in or still work on.
Miminho
Full Stack Developer
Handmade jewelry store with a tailored admin: products, orders, and content in one place - less day-to-day chaos. Clear returns/complaints, SEO, fast storefront. Next.js + Strapi.
LeasePulse
Frontend Engineer
Working with a team on a financial services organization's leasing operations dashboard with AI-powered document extraction. Overall the initiative contributed to roughly 30% faster approvals and higher accuracy on contracts. My contribution is mainly frontend: status monitoring, multi-step approvals, exception handling, API integrations, plus collaboration on annotation tooling and Azure ML-related parts of the product.
MDT Net
Frontend Engineer
Collaborating with a team on large-scale healthcare software: patient portal (appointments, records, prescriptions, messaging) and call-center panels - in practice faster handling at thousands of daily interactions and fewer tool switches for staff. My work focused mainly on the frontend - React, API integrations - within HIPAA-aware requirements.
Workflows
Frontend Engineer
Internal platform for teams processing millions of text documents at a time. Instead of gluing scripts together for every change - a visual flow builder where you arrange steps like a diagram, live dashboards for throughput and data quality (D3.js), and many people tagging in parallel with WebSocket sync. It streamlined collecting, filtering, and preparing data for ML models and reporting - a shorter path from raw text to an export-ready dataset. React, NLP (WinkNLP), AWS.
Protaggy
Frontend Engineer
Platform for building ML training datasets across text, video, and images. It was deployed at universities in the United States, among other settings. In practice hundreds of annotators could work at once - labels, bounding boxes, segmentation, classification - with label-quality views and model-metric dashboards. Dataset versioning, roles, project assignments, and export into training pipelines: less chaos than spreadsheets and email attachments, faster iterations between model versions, with thousands of training samples moving through the system. React, WebSockets, D3.js, AWS.