Speech & LLM algorithms
LLM supervised fine-tuning (SFT), LLM-based speech generation, prosody and stress modeling for expressive TTS, NLP for the TTS front end, and on-device spoken language identification.
About me
I am an M.S. student in Artificial Intelligence at the University of Southern California, graduating in May 2027. Before USC, I spent six years at Mobvoi, growing from speech algorithm engineer to tech lead. I shipped multilingual speech and NLP models to tens of millions of users, led speech-LLM services and LLM agent frameworks, and built the data and evaluation pipelines behind them.
At USC, I research computer-use and research agents: how they plan, where they break, and how to make them recover. I’m broadly interested in LLM agents, speech and NLP, multimodal learning, and human–AI collaboration.
Now, as an AI Engineer Intern at Wyze, I build AI that listens, speaks, and acts, running in real time on home cameras.
I’m looking for PhD positions and full-time ML/AI engineering roles starting in 2027. Read my CV or email me.
What I work on
Multimodal agents
Agentic systems for research and production: how they plan, where execution breaks, and how they recover safely.
Data & evaluation
Multimodal data pipelines and evaluation systems that tell you whether a model is actually ready to ship.
Side project, built solo
Audio Story Visualizer
An AI-powered audio visualization and dubbing platform that turns music, podcasts, and audiobooks into editable videos with high-fidelity speech processing.
Try the demo at qinzhi.org
Experience
-
Jul 2026 – now
AI Engineer Intern, Wyze
Developing and integrating core components for real-time conversational audio agents on edge devices, including model integration, automated testing, language identification, and speech-to-intent optimization.
-
Oct 2025 – May 2026
Research Assistant, University of Southern California
Researched computer-use and research agents, with a focus on tool-use planning, execution robustness, and failure modes.
-
Aug 2025 – Mar 2026
Part-time Intern, Mobvoi
Optimized production speech models and LLM-based multi-agent workflows, and supported data pipelines and system integration.
-
May 2023 – Jul 2025
Tech Lead, Mobvoi
Led production LLM agent frameworks, speech-LLM models and services, multimodal data engineering, and model evaluation systems; mentored junior engineers and interns.
-
Jul 2019 – May 2023
Speech Algorithm Engineer, Mobvoi
Built multilingual speech and NLP models, scalable backend services, and standardized data pipelines for production systems serving tens of millions of users.
-
Aug 2018 – Dec 2018
Research Intern, TAL Education Group
Developed graph-enhanced deep knowledge tracing models and co-authored work published at AIED 2019.
Education
-
2025 – 2027
M.S. in Artificial Intelligence, University of Southern California
Expected May 2027.
-
2016 – 2019
M.S. in Software Engineering, Beijing University of Technology
-
2012 – 2016
B.E. in Computer Science, Southwest Minzu University
Ranked top 5 of 154 students.
Publications
Google Scholar-
arXiv 2025
-
Applied Intelligence 2024
-
APSIPA ASC 2023
-
PRML 2022
-
AIED 2019
-
JPCS 2019
-
CCIOT 2018
-
CCIS 2018