
About Me
I’m an AI safety researcher, a software engineer based in New York.
I engage with AI safety since 2018, spanning from AI policy to AI safety technical work. My research focus is how AI perceives its human user, and I use mechanistic interpretability. I investigated ML-application in tenant-screening, and forecasted Arroyo v. CoreLogic (2018). My ML policy work was accepted by ACM FAccT 2019.
I’m particularly good at cross-domain pattern recognition and connecting dots that most didn’t see. My best work has been constantly achieved by reframing an intractable problem into one with feasible solutions. I thrive in a lot of intellectual complexity and with execution autonomy.
I’m actively looking for collaborators! I’d be happy to hear about your work, let’s chat!
Selected Work
- I and Thou: Turning the Mirror on the Machine · Feb 2026
- Llama2 Safety Evaluation Anatomy · Jan 2026
- LLM Hallucinations: An Internal Tug of War · Aug 2025
My Experience
Before moving to the US, I was part of the founding product and marketing team at Smart Order in Hangzhou, China. I moved to NY to study data science in NYU CUSP. After that, I spent a year as a research fellow with the New York City Commission On Human Rights, focusing on AI policy and fairness in machine learning. I play Guqin (a Chinese zither). I also practise Chinese calligraphy.
Contact
- Email: violazhongg[at]gmail.com
- GitHub: github.com/violazhong
- X x.com/viola_zhongg
- Substack substack.com/@violazhong