Wei Zhang

I received my Ph.D. from the University of Wisconsin–Madison, where my dissertation focused on improving the reliability of concurrent software through an effect-oriented approach.

I joined IBM Research as a Research Staff Member in August 2013. Since then, my research has spanned programming languages, distributed deep learning, systems for machine learning, and AI for Code. My publication record is available on Google Scholar: https://scholar.google.com/citations?user=DJMSA3YAAAAJ&hl=en

Some of my research highlights include:

1. Decentralized distributed deep learning, including work published at NeurIPS 2017 and ICML 2018.
2. Speech workload acceleration, where our work helped reduce AI speech recognition training time from one week to 11 hours: https://venturebeat.com/technology/new-ibm-technique-cuts-ai-speech-recognition-training-time-from-a-week-to-11-hours
3. COBOL-to-Java modernization using large language models, contributing to IBM watsonx-based generative AI capabilities for mainframe application modernization: https://venturebeat.com/ai/ibm-taps-watsonx-generative-ai-to-help-modernize-cobol-on-mainframes

At IBM Research, I have worked across several research groups:

• August 2013 – July 2015: X10 Group
• July 2015 – January 2016: Programming Languages Group
• January 2016 – September 2016: System Analysis and Optimization Group
• September 2016 – September 2021: System Acceleration for Machine Learning Group
• September 2021 – Present: AI for Code Group