Jinhao Zhao

researcher @ Zhuiyi Technology

About

I am a Ph.D. student in Computer Science at the McKelvey School of Engineering, Washington University in St. Louis. My research focuses on theoretical computer science, including parallel algorithms, distributed systems, load balancing, and real-time scheduling. I have published papers in top conferences such as SPAA, PPoPP, RTSS, and ECRTS, with one of his works receiving the Best Paper Award at SPAA 2024. Before pursuing my doctoral studies, I studied in yaoclass of Tsinghua University, then worked at Zhuiyi.ai on industrial automatic speech recognition and at Ruyi.ai on knowledge graph–based risk detection and recommendation systems. My broader interests span algorithm design, machine learning, and large-scale system optimization. Outside of research, I am an active bridge player, having placed 2nd in the North American Collegiate Bridge competition. I also created open-source projects with over a thousand active users.

Country

United States

City

Greater St. Louis

Industry

Information Technology & Services

Skill

Kaldi, Chatbot Development, 聊天机器人, Computer Vision, 计算机视觉, Academic Research, Object-Oriented Programming (OOP), Machine Learning, 机器学习算法, 面向对象语言, 学术研究

Experience

researcher

Zhuiyi Technology

2019-11 - 2021-8 · 1 yr 10 mos

中国广东省深圳

Trained and optimized the ASR model (based on Kaldi, in Python and C++) according to data from industrial environment. Applying the technology of appending new phrases to the lattice artificially, it reached 98%+ accuracy in technical dialogs. Applied high performance computing methods, reducing the lag from 500+ms to <200ms.

researcher

海知智能

2019-2 - 2019-8 · 7 mos

中国北京市海淀区

Built the risk detection model for companies by knowledge graph, and the recommendation system for users with financial products. The outputs were designed to be understood by human experts, and was applied practically in China CITIC Bank and Bank of China.

researcher

MEGVII FaceID

2016-8 - 2016-12 · 5 mos

中国北京市海淀区

Conducted for some deep learning methods in computer vision, including tracking, re-identification, body parsing and human attributes. The result in re-identification was once the SOTA and summarized as an intent for the company.