Nanshu Wang
Staff Software Engineer at Meta
California, United States
Summary
👤
Senior🎓
Top SchoolNanshu Wang is a Staff Software Engineer at Meta with 12 years of experience specializing in LLMs, NLP, and on-device AI for conversational systems. She leads work across post-training for large language models, tool use and retrieval-augmented generation, multimodal NLG, privacy-aware ML, and integrity for smart communications. Her hands-on contributions include ML engineering in the popular PyText framework—adding knowledge distillation, document classification losses, dense feature support, and byte-token inputs to improve production NLP pipelines. Trained across Renmin University, UCAS, and Carnegie Mellon, she blends rigorous academic foundations with practical system-building from speech recognition to large-scale conversational AI. Colleagues rely on her to bridge research advances and deployable ML solutions that respect user privacy and device constraints.
13 years of coding experience
4 years of employment as a software developer
Master, Computer Science, Master, Computer Science at University of Chinese Academy of SciencesBachelor’s Degree, Bachelor’s Degree at Renmin University of ChinaMaster, Computer Software Engineering, Master, Computer Software Engineering at Carnegie Mellon University
English, Chinese, Korean
Github Skills (9)
pytorch10
machine-learning10
distill10
nlp10
dis10
natural-language-processing10
python9
onnx9
tensorflow4
Programming languages (3)
Github contributions (5)
facebookresearch/pytext
Jan 2019 - May 2021
A natural language modeling framework based on PyTorch
Role in this project:
ML Engineer
ML EngineerContributions:14 commits, 16 PRs in 2 years 4 months
Contributions summary:Nanshu primarily focused on implementing and improving machine-learning related functionalities within the PyText framework. They contributed to knowledge distillation techniques, creating new tasks and loss functions for document classification, as well as onnxable lasttimestep pooling. The user also added support for dense features for seqNN models, and made enhancements to support byte token input for language models. They also fixed issues related to GPU inference and provided modifications for the exporter.
language-modelingpytorch
gardenia22/gardenia22.github.io
Jan 2015 - Feb 2020
Contributions:148 commits, 148 pushes in 5 years 1 month
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.