Microsoft
Jul 2025 – Present- Develop LLM agent workflows involving memory, reasoning, actions, and reflection for AI applications.
- Develop presentation-generation workflows and evaluate outputs for AI applications.
Professional résumé
Jinghua Yao, Ph.D.
Principal Applied Scientist · Generative AI & Machine Learning
Read my full résumé below. A PDF copy is available on request.
Professional summary
Applied scientist with experience at Microsoft, Apple, and Samsung, combining a Ph.D. in mathematics with hands-on development of enterprise AI systems. Expertise spans LLM agents, distributed fine-tuning, multimodal learning, NLP, time-series forecasting, and computer vision. Track record of architecting applications, automating operational workflows, and translating mathematical modeling into practical machine learning solutions.
Technical expertise
- Generative AI
- LLM agents, LLM/VLM fine-tuning, transformers, prompt engineering, document Q&A, summarization, multimodal understanding, embeddings
- Machine learning
- NLP, sentiment analysis, forecasting, anomaly detection, CNNs, ensemble methods, statistical modeling, optimization
- Engineering
- Python, PySpark, C#, TypeScript, Go, scikit-learn, Docker, AWS, Amazon S3, Splunk, Git, Streamlit, Gradio
Professional experience
Apple
May 2020 – Jun 2025- Architected and implemented visual understanding with late interaction for enterprise applications.
- Developed distributed fine-tuning workflows for LLMs and vision-language models, including Mistral, Florence-2, Phi-3.5, and Molmo, supporting embedding, Apple Support, captioning, and generative AI applications.
- Built enterprise LLM applications for customer sentiment analysis, contextual Q&A, document analysis, and summarization, improving feedback resolution and reducing labor and device costs.
- Architected and implemented automated survey-health reporting using Splunk, Amazon S3, anomaly detection, and email notifications, replacing manual team workflows.
- Built time-series forecasting, trend analysis, and anomaly detection systems for call centers and survey services; supported improvements in call-answer rates.
- Developed transformer-based sentiment analysis and classification for customer comments.
- Delivered ML and AI technical lectures as a main contributor to a series of 32+ one-hour sessions on modeling and algorithms.
Samsung Display American Lab
Mar 2019 – May 2020- Led problem formulation and development of a display-defect detection and classification system used in products.
- Combined convolutional neural networks, dimensionality reduction, and classical ML using approximately 800 MAP and TEM images; developed a product-used classification solution.
Research & academic experience
Kent State University
Sep 2016 – Sep 2017The University of Iowa
Aug 2013 – Jan 2016Research in nonlinear analysis, partial differential equations, dynamical systems, and variational methods, with journal publications in Physica D, Journal of Differential Equations, and Communications in Mathematical Sciences.
Selected publications
2019
Patterns of complex oscillations and instability in chemical reactions
Communications in Mathematical Sciences · 17(6), 1713–1736
2017
Spectrum and bifurcation for semilinear elliptic problems in ℝᴺ
Journal of Differential Equations · 263(9), 5939–5967
2014
O(2)-Hopf bifurcation for a model of cellular shock instability
Physica D: Nonlinear Phenomena · 269, 63–75
Education
Ph.D. in MathematicsIndiana University Bloomington, USA
Master’s in MathematicsLanzhou University, China
Bachelor’s in MathematicsLanzhou University, China
Professional service & recognition
Invited reviewer for Mathematical Reviews, American Mathematical Society; journal referee for Scientific Reports, Journal of Mathematical Analysis and Applications, and other mathematical journals.
National Prize and Provincial Special Prize, National Mathematical Modeling Contest, 2004.
Request a PDF copy
For recruiting or professional correspondence, please contact jinghuayao AT gmail DOT com.
Back to homepage ←