I build mobile security SDKs for Android and iOS. My work spans native development, platform compatibility, and security detection.
I also build tools that make working with AI agents easier to understand and review.
I’m focused on reducing AI-generated code slop through small, reviewable steps and understanding the core logic before moving on. I built Baby Steps to make that understanding part of the development workflow.
I’m also exploring code reviews grounded in team rules and conventions, system context, and change history. Those details matter to me when building reliable products and keeping software maintainable over time.
- Korean Humanizer — An agent skill for making Korean writing more natural while preserving its meaning
- DocPilot — A desktop Markdown editor with Claude and Codex sessions, where you can review file changes before accepting them
- Kev Review Gate — An experimental local code-review triage model adapted from Kev, with reproducible training, synthetic evaluations, and Claude/Codex agent skills.
C/C++ · Kotlin · Java · JNI/NDK · Objective-C · Swift
Small changes. Clear explanations. Tests and diffs I can inspect.