I'm Mengya Hu, a Senior Applied Scientist at Microsoft working on Responsible AI; in English settings I go by Mia Hu. My focus is LLM content safety — moderation systems, hallucination detection and mitigation, and the design trade-offs between safety and user experience.
Before that, I did my Ph.D. in Mechanical and Aerospace Engineering at Princeton, working on exoplanet detection — building simulation pipelines for a starshade–telescope system and designing detection methods that pull faint signals out of images.
I did my undergrad at USTC, in Thermal Science and Energy Engineering.
Hunting planets and moderating AI sound worlds apart, but to me they're the same problem: reliably detecting the signal you care about inside overwhelming noise — once a dim planet beside a star's glare, now a harmful output or a hallucinated claim inside oceans of generated text.
I'm endlessly curious about the world, with a broad interest in frontier science and technology. This blog focuses on frontier AI and industry trends — with occasional forays into other fields.
Every post is published in both English and Chinese.
Recent writing:the AI layoff wave isn't in the data,what happens to HIPAA when your records meet ChatGPT, and the Kimi K3 audit gap.
All opinions here are my own and do not represent my employer.