Frontend that makes datasets exceeding 500k rows viewable in real time, directly in the browser, without crashing or degrading the user experience. Built infinite scroll and dynamic loading to render the dataset, optimized parsing and state management to avoid UI blocking, and maintained smooth interaction under continuous data load.
Corporate governance platform for managing assemblies, discussions, and voting, enhanced with AI for real-time transcription, summarization, and translation. Includes a meeting assistant with AI listening and live chat, a full video call system integrated alongside the AI and chat layers, and a PDF-to-assembly generation feature that auto-extracts shareholder meeting data.
Personal project for tracking income, expenses, and budgets in a single mobile app, built end-to-end from design to release.
Scalable component library to standardize UI development and accelerate delivery across applications, with reusable components (lists, loaders, skeletons, feedback states) built on standardized visual and interaction patterns with accessibility in mind, and a structure that handles every UI state: loading, empty, error, and success.
Home security platform with live WebSocket updates, used by 100+ concurrent call-center agents monitoring doors, gates, and alarm status in real time. Built a real-time UI with live status for multiple devices, responsive dashboards for continuous monitoring, and stable connections under high-frequency updates.
Bundle size audit and cleanup for the platform's core Vite application, targeting load performance. Audited and removed barrel-file re-exports across the codebase, re-structured module imports to enable proper tree-shaking, and validated the gains with bundle analysis before and after the change.
Desktop app that reads veterinary ultrasound images and extracts structured metrics (measurements, tissue observations, diagnostic notes), then exports a clean report. Built with Electron for local file processing and OpenAI Vision API integration with structured prompt engineering, automating extraction into an exportable report.
Ported the Windows-only game to run natively on macOS, without relying on Windows emulation. Diagnosed a Java environment check that only validated Windows-style paths, then built a cross-platform runtime handler that checks macOS install locations and PATH before falling back to the original Windows logic, adapting launch and file-handling so the game runs natively.