I build practical software products at the intersection of frontend systems, native macOS, and AI-assisted engineering.
Previously, I worked on high-scale consumer and internal platforms at TikTok, NetEase and Alibaba. Recently, I have been building independent products with a more AI-native workflow, using tools like Codex and Claude Code for research, implementation, debugging, testing, and documentation.
- Apple-powered live caption apps: real-time meeting, livestream, audio, and video captioning powered by Apple's on-device ASR models, with local-first transcripts, playback, export, diagnostics, and release packaging.
- Platform systems: low-code tools, micro-frontends, resource management workflows, e-commerce flows, and live streaming interaction systems.
- 6.1B+ coding-agent tokens in one month across Codex and Claude Code, used for architecture, implementation, debugging, testing, and documentation.
- LeetCode confession: before joining TikTok, I grinded a lot of LeetCode. I have forgotten most of those patterns now, but years of building real top-tier products made me a better engineer than my old interview-prep self.
- With a software engineering background and 10+ years of hands-on product engineering, AI coding agents let me learn unfamiliar domains quickly, turn unknown tasks into working systems, and still bring human judgment from production-scale experience.
- I treat AI coding agents as engineering infrastructure: fast iteration, deep codebase reading, test-driven verification, and careful human review.
- Product-minded frontend and full-stack engineering.
- React, Node.js, Swift, SwiftUI/AppKit, and macOS app development.
- Performance, diagnostics, maintainable architecture, and useful internal tools.