About Me
Bio
Applied Scientist at Amazon, building LLM-based agentic reasoning systems and adaptive decision pipelines, with prior research in theoretical machine learning: identifiable representation learning, signal processing, matrix factorization, and non-convex optimization.
I received my Ph.D. from the Department of Computer & Information Science & Engineering (CISE) at the University of Florida, advised by Prof. Kejun Huang. Before joining UF, I obtained my bachelor’s degree from Nanjing University.
Research Focus
Machine Learning Foundations
Representation learning with identifiability, non-convex optimization, and signal-processing-grounded latent variable modeling.
Adaptive Decision Systems
Contextual bandits, partial-feedback learning, and uncertainty-aware decision making pipelines for ML system at scale.
LLM & Agentic Reasoning Systems
Routing, orchestration, and control strategies for multi-component reasoning systems built on large language models.
Selected Research & System Contributions
Published Research
- Established identifiability results for latent representation models, including bounded component analysis and dictionary learning.
- Developed non-convex optimization methods with theoretical guarantees fo identifiable latent representation learning problems.
Applied/System Work at Industry
- Built uncertainty-aware decision methods for recommendation settings with partial feedback.
- Developed routing and orchestration strategies for LLM-based agentic reasoning systems.
- Focused on system-level control design and decision quality under practical constraints.
Selected Publications
Global Identifiability of L1-based Dictionary Learning via Matrix Volume Optimization
Establishes global identifiability guarantees for L1-based dictionary learning via a matrix-volume formulation.
Identifiable Bounded Component Analysis via Minimum Volume Enclosing Parallelotope
Introduces a minimum-volume enclosing parallelotope view for identifiable bounded component analysis.
Complex Bounded Component Analysis: Identifiability and Algorithm
Provides identifiability analysis and an efficient algorithm for complex-valued bounded component analysis.
Academic Service
- Conference reviewer: NeurIPS, ICML, ICLR, AISTATS, AAAI, ICASSP, MLSP, IJCNN.
- Journal reviewer: IEEE Transactions on Signal Processing (TSP), Journal of Machine Learning Research (JMLR).
Links
Previous Research Keywords (kept for continuity)
Machine Learning, Representation Learning, Unsupervised Learning, Optimization, Latent Variable Models, Non-convex Optimization, Uncertainty Estimation, Recommendation Systems, Reinforcement Learning, Large Language Models, Agentic System.