Reliable & Secure AI Systems

Wei Hao

I build the systems that make AI trustworthy by enterprises in the real world — reliable when it serves, and secure when it's attacked.

Founder of an AI startup company PhD, Computer Science · Columbia University New York City
Read the research View CV
Portrait of Wei Hao
01 — Vision

Make AI systems we can actually depend on.

I received my PhD degree from Columbia University. My PhD work spans the full lifecycle of deploying machine learning in production — from serving and monitoring models at scale to defending them against adversaries and detecting their misuse. The thread connecting it all: AI is only useful when it is reliable, observable, and safe. My systems are deployed in companies including Microsoft and Barracuda Networks.

Co-advised by Asaf Cidon & Junfeng Yang.

ML Systems at Scale

Predictable serving, model versioning, and on-device monitoring — Clockwork (OSDI), MGit (ICML), Nazar (ASPLOS).

AI Security & Robustness

Evasive attacks on edge models, jailbreak analysis, and robust deepfake-audio detection.

Detecting AI Misuse

Generalized LLM-generated text detection and large-scale study of LLM-generated malicious email.

02 — Timeline

Recent news

Present
Building an AI startup company.
May 2026
Officially received my PhD degree from Columbia University.
Apr 2026
Honored to receive the Davide Giri Memorial Prize for combining research excellence with efforts to promote research collaboration.
Apr 2026
My dissertation received the G-Research PhD Prize.
Jun 2025
Our study on the prevalence of LLM-generated malicious emails was covered in Forbes and TechRepublic.
May 2025
Learning to Rewrite: Generalized LLM-Generated Text Detection accepted to the main conference of ACL 2025.
Mar 2025
Do Spammers Dream of Electric Sheep? — characterizing LLM-generated malicious emails — accepted to ACM IMC 2025.
Jan 2025
Diversity Helps Jailbreak Large Language Models accepted to NAACL 2025.
Jan 2025
I Can Hear You: Selective Robust Training for Deepfake Audio Detection accepted to ICLR 2025.
03 — Selected Work

Publications that shaped the thesis

All publications →
04 — Honors

Awards & recognition

2026
Davide Giri Memorial Prize
Given to a Computer Science PhD who combines excellence in research results with continued outstanding efforts to promote research collaboration.
Columbia CS
2026
G-Research PhD Prize
Awarded for outstanding doctoral research.
Dissertation
2023
ML and Systems Rising Star
Recognized among the field's emerging researchers at the intersection of machine learning and systems.
MLCommons
2022
MLSys Student Travel Grant
MLSys
2020
OSDI Distinguished Artifact Award
For Clockwork — a distributed DNN serving system with predictable end-to-end performance.
OSDI