I design and ship software across AI applications, developer tools, Windows system utilities, embedded devices, and interactive web experiences.
My focus is not technology for its own sake. I care about turning complex workflows into products that are clear to use, reliable in operation, and maintainable over time—from architecture and implementation to interface design, documentation, and release.
I build the complete path from a real problem to a dependable product.
|
A developer tool that brings Claude Code workflows, Agent Teams, and Subagents to third-party models such as DeepSeek, GLM, Kimi, and Qwen.
Focus: model integration, workflow orchestration, and practical AI tooling. |
A local-first workspace that combines natural-language analysis, code execution, semantic modeling, visualization, and report generation.
Focus: AI product architecture, data governance, and full-stack integration. |
|
A Go and Vue 3 system tool for managing sing-box, Xray, mihomo, and Hysteria2 through one consistent workflow.
Focus: configuration orchestration, protocol compatibility, and desktop product design. |
A single-file atmospheric sonar maze where sound becomes the map, featuring four chapters, local co-op, and spatial audio.
Focus: interaction design, real-time audio, and creative browser experiences. |
WeightScale · Embedded-to-web product system
A complete hardware-to-browser workflow built across two repositories:
| Device firmware | Browser dashboard |
|---|---|
| WeightScale-workspace — STM32F103 firmware for HX711 weighing, temperature sensing, display control, touch interaction, Bluetooth, and serial protocols. | WeightScale-web — Web Serial dashboard for live weight, waveform, statistics, logs, and device events. |
Sensors → STM32 firmware → USB / Bluetooth serial → Web Serial → Browser dashboard
Focus: embedded acquisition, protocol design, hardware–software integration, and real-time visualization.
| Area | Strengths |
|---|---|
| AI engineering | Agent workflows, model endpoints, data analysis systems, long-running task recovery |
| Backend and systems | Go services, process management, configuration generation, HTTP APIs, SQLite |
| Frontend and product | TypeScript, Vue, React, Vite, responsive interfaces, data visualization |
| Embedded and devices | STM32, C, sensors, serial protocols, device-to-browser integration |
| Delivery | Architecture, observability, documentation, reproducible setup, release workflows |
- Start with the workflow. Understand the outcome before choosing the abstraction.
- Make systems observable. Expose state, logs, errors, and useful feedback.
- Design for recovery. Treat interruption, invalid input, and failure as normal cases.
- Ship the complete path. Implementation, interface, documentation, and release all matter.
Build useful things. Make them dependable.