About Me
I’m Yi (Justice) Zhang, a production ML/NLP engineer with more than two years of experience shipping multilingual LLM and NLP systems at scale. I am currently pursuing a Master of Data Science in Computational Linguistics at the University of British Columbia in Vancouver, BC.
At BIGO, I built and operated Trust & Safety systems that analyzed text transcripts generated from audio. My work covered data preparation, supervised fine-tuning (SFT), quantitative evaluation, model serving, production integration, staged rollout, threshold calibration, and human-in-the-loop workflows.
My research includes function calling at ICASSP 2025 and multi-agent collaboration at Findings of EMNLP 2024. I am interested in LLM engineering and agentic AI applications.
Education
- University of British Columbia, Master of Data Science in Computational Linguistics — September 2026–June 2027 (expected)
- Beihang University, M.Eng. in Computer Technology — September 2021–June 2024
- Chongqing University, B.Eng. in Software Engineering — September 2017–June 2021
Experience
BIGO — Machine Learning / LLM Engineer
July 2024–August 2026 · Guangzhou, China
- Shipped and operated multilingual LLM and BERT moderation systems for IMO and Bigo Live across six languages.
- Built a product-aware LLM baseline from 12.5M+ human-review examples across two products, preserving hard negatives and label provenance and calibrating per-language thresholds.
- Raised confirmed violations by 55% at constant review volume for the first three migrated languages while consolidating the six-language fleet and cutting GPU footprint by 72%.
- Delivered an internal RAG and semantic-search tool that let investigators retrieve similar cases, filter by metadata, and reconstruct surrounding room context.
BIGO — Machine Learning Engineer Intern (Speech)
December 2023–June 2024 · Beijing, China
- Fine-tuned and deployed an Arabic neural TTS model for BIGO Live, then built a multilingual TTS service with timestamped subtitles for Likee; both reached users after product or integration testing.
Microsoft Research Asia — Research Intern
December 2021–October 2022 · Beijing, China
- Designed and implemented Ada-TISE, a dual-branch Transformer for text-informed speech enhancement under unreliable alignment.
- Improved PESQ, CSIG, CBAK, and COVL by 0.12, 0.11, 0.05, and 0.11 over a FLOPs-balanced baseline on Voice Bank+DEMAND under the low-SNR condition.
Open to Opportunities
- During the program: Open to sustained part-time ML engineering roles in Vancouver or remotely in Canada, up to 24 hours per week, and full-time remote roles with employers outside Canada.
- From July 2027: Seeking full-time and new-graduate ML engineering roles.
- Interests: NLP and LLM engineering, production ML, Trust & Safety, and agentic AI applications.
Publications
- Wei Zhang*, Yi Zhang*, Li Zhu, Qianghuai Jia, Feijun Jiang, Hongcheng Guo, Zhoujun Li, Mengping Zhou. ADC: Enhancing Function Calling Via Adversarial Datasets and Code Line-Level Feedback. ICASSP 2025. [paper]
- Wei Zhang, Hongcheng Guo, Jian Yang, Zhoujin Tian, Yi Zhang, Chaoran Yan, Zhoujun Li, Tongliang Li, Xu Shi, Liangfan Zheng, Bo Zhang. mABC: Multi-Agent Blockchain-inspired Collaboration for Root Cause Analysis in Micro-Services Architecture. Findings of EMNLP 2024. [paper]
- Wei Zhang, Xianfu Cheng, Yi Zhang, Jian Yang, Hongcheng Guo, Zhoujun Li, Xiaolin Yin, Xiangyuan Guan, Xu Shi, Liangfan Zheng, Bo Zhang. ECLIPSE: Semantic Entropy-LCS for Cross-Lingual Industrial Log Parsing. arXiv, 2024. [paper]
- Yi Zhang, Zhiyuan Zhao, Dacheng Yin, Yucheng Zhao, Chuanxin Tang, Chong Luo. Ada-TISE: Adaptive Text Informed Speech Enhancement. Technical Report, 2022.
Contact
Last updated: September 12, 2026.