Hanlin Zhao (Zhao Hanlin)

I lead Z.ai AutoTyper (Zhipu AI Input Method) and the GLM-ASR / GLM-ASR-Nano speech model lines.

What I build is specific: a voice-first writing workflow that lives inside any input field. Users can speak to draft, edit in-place (rewrite, translation, tone/style), and hand off tasks to agent execution (planning + tool-calling) when needed. The most important part is writeback: results return to the same context the user is writing, not to a separate chat window.

I also invest in robustness and evaluation: real-world speech (dialects, quiet speech, noise, mixed languages) and benchmarks that reflect natural prompts. GLM-ASR-Nano-2512 has 502,953 all-time downloads on Hugging Face (as of 2026-02-01).

Earlier at Z.ai: NaturalCodeBench (ACL 2024 Findings). I also co-authored work on GUI agents and evaluation: AutoGLM, VisualAgentBench, and the ChatGLM report.

If you're exploring speech input, on-device inference, agent workflows, or benchmark design, feel free to reach out.

:mailbox: hanlin9908@gmail.com · :mortar_board: Google Scholar