Pin-zu Li
LLM Engineer at Taiwan AILabs
Taiwan
Summary
👤
Senior🎓
Top SchoolPin-zu Li is an LLM engineer with eight years of experience building production-grade language models and NLP systems in Taiwan, combining academic research from NTU/Academia Sinica with industry delivery. He has led domain-specialized LLM efforts—training a financial LLM and inventing a cost-effective fusion of finance and Taiwan-specific models—while co-designing a multimodal RAG system that improved zh-TW benchmark performance by 7%. His research background includes novel fake-news early-detection methods that generate domain-invariant and reference features via soft prompts, and he helped develop TAIDE, a Traditional Chinese LLM with GPT-3.5-comparable performance and a practical “chat vector” technique for transferring instruction-following behavior. Comfortable spanning data pipelines, retriever training, evaluation engineering, and post-training optimization, he blends cultural-linguistic curation with scalable modeling to solve real-world cross-lingual and domain-adaptive challenges.
8 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering - BE, Computer Science and Information Engineering, Bachelor of Engineering - BE, Computer Science and Information Engineering at 國立中正大學Master of Science - MS, Data Science, Master of Science - MS, Data Science at 國立臺灣大學
Github Skills (27)
pytorch10
python10
llama10
machine-learning10
inference10
llm10
deep-learning10
tensorflow10
natural-language-processing10
bert10
cuda10
transformer10
speech-recognition10
nlp10
transformers9
Programming languages (3)
Github contributions (5)
aqweteddy/BERTDistilForClassification
Jun 2020 - Sep 2022
An Unofficial Implementation for Distilling Task-Specific Knowledge from BERT into Simple Neural Networks
Contributions:6 commits, 3 PRs, 1 branch in 2 years 3 months
nlpdistillingbertknowledgeneural-networks
aqweteddy/TextGeneration
Jan 2021 - Mar 2021
Contributions:27 commits, 8 pushes, 1 branch in 2 months
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.