Projects

A visual-first AI tool discovery and workflow orchestration engine powered by a dual-stage pgvector semantic search and parallel Gemini-Flash live discovery pipeline. It maps unstructured natural language queries to high-dimensional embeddings and synthesizes interactive, node-based pipelines on an infinite React Flow canvas.

A robust, concurrent testing framework for evaluating AI agents and scripts using multi-tiered assertions and LLM-as-a-judge.

A high-performance, frameless Electron and React productivity dock for Windows that orchestrates system-level tools, secure terminals, and AI workflows through a sandboxed, low-latency desktop overlay.

An open-source forked project of my original Float Dock codebase which I actively co-developed, DockShift is a highly integrated, performance-optimized command dock for Windows that collapses complete desktop workspaces, GPU-rendered terminals, and sandboxed AI assistants into a single global hotkey. It hooks directly into the Win32 subsystem to capture and restore full application layouts, eliminating context-switching latency while maintaining strict security boundaries.

Envoy-powered LLM traffic gateway with dynamic provider routing, transparent fallback, and local E2E tests.

A high-performance, offline-first 75-day challenge engine and habit tracker built on React 19 and Capacitor, featuring transactional bi-directional cloud synchronization. Engineered with an asynchronous Supabase sync layer, the application guarantees zero-latency user interactions by offloading remote database operations from the UI thread via local storage state mirroring.

AI-powered developer telemetry platform that analyzes coding activity to generate insights on productivity, skill growth, debugging patterns, and career progression.