At Huawei, I reduced peak memory by 30–50% on large projects through index segmentation and on-demand unloading, while improving project initialization and update performance. I worked on Python IDE internals and exposed code intelligence to coding agents through MCP symbol and workspace tools.
I built engineering tools including a Java annotation processor and a Gradle plugin for CI test reports, profiling artifacts, and benchmark regression analysis. I also co-designed and shipped an internal VS Code extension for running coding agents on software tasks, and built a benchmark runner to compare agents.
In my open-source work, I built OpenAIDE, a workbench for running coding agents over the Agent Client Protocol, with a Rust application server and shared TypeScript frontend. I’ve also contributed merged pull requests to the Agent Client Protocol, its Codex adapter, and rust-lang/futures-rs.
