CV

Updated Mar 2026

Basics

Name Yani Guan
Label Ph.D., UCLA | ML for Science · GNN · LLM · Generative Models
Email yaniguan@ucla.edu
Phone 310-853-9631
Url https://yaniguan.github.io
Summary Ph.D. in Chemical Engineering from UCLA (December 2026), authorized to work in the U.S. (OPT) and will not require visa sponsorship now or in the future, building production ML systems at the intersection of machine learning and molecular science. Research spans end-to-end ML pipelines, GNNs, transformer and diffusion-based generative models, and LLM-powered agentic platforms for scientific automation.

Work

  • 2024.01 - now
    Graduate Student Researcher — End-to-End ML Systems
    University of California, Los Angeles
    Model development and production ML pipelines for molecular science.
    • Designed and deployed end-to-end ML pipelines spanning data cleaning, feature engineering, model training, evaluation, and iterative redeployment
    • Built GNN and transformer-based models trained on large structured datasets to predict complex molecular properties
    • Developed ML interatomic potentials as high-fidelity surrogates for physics-based simulations with focus on memory efficiency and OOD generalization
    • Engineered physics-informed features and uncertainty quantification methods to improve model robustness
  • 2023.09 - now
    Graduate Student Researcher — ChatDFT (LLM Agentic Platform)
    University of California, Los Angeles
    LLM-powered agentic simulation platform for quantum chemistry automation.
    • Architected an AI agent system integrating LLMs with RAG, intent extraction, and HPC job scheduling to fully automate quantum chemistry workflows — achieving >70% reduction in setup-to-results time
    • Shipped production-grade agentic pipelines (Python, REST APIs, cloud infrastructure) with structured evaluation loops for continuous model monitoring
    • Applied SFT and RLHF to fine-tune domain-specific LLMs for molecular modeling and data interpretation
  • 2022.09 - now
    Graduate Student Researcher — Data-Driven Molecular Design
    University of California, Los Angeles
    Structure–activity relationships and generative approaches for molecular discovery.
    • Applied data-driven ML methods to analyze structure–reactivity relationships across broad chemical libraries
    • Designed generative model workflows combining diffusion-based and graph-based approaches to explore chemical space
    • Contributed to multiple peer-reviewed publications on electrochemical amination and ML for heterogeneous catalysis
  • 2022.06 - 2022.09
    Algorithm Researcher — ML Infrastructure & Production Deployment
    DP Technology
    ML infrastructure and production deployment for neural network potentials.
    • Built and shipped production ML pipelines integrating neural network potentials with molecular dynamics engines (Python, C++), deployed on cloud HPC infrastructure
    • Developed robust, reusable software tools adopted by 200+ researchers across academia and industry

Education

  • 2022.09 - 2024.03

    Los Angeles, CA

    M.S.
    University of California, Los Angeles
    Chemical Engineering
  • 2022.09 - 2026.12

    Los Angeles, CA

    Ph.D.
    University of California, Los Angeles
    Chemical Engineering
  • 2018.09 - 2022.06

    Tianjin, China

    B.S.
    Hebei University of Technology
    Chemical Engineering

Awards

Certificates

Python for Everybody Specialization
University of Michigan 2024-05-01

Skills

ML Frameworks & Models
PyTorch
scikit-learn
HuggingFace Transformers
Graph Neural Networks (GNN)
Transformer Architectures
Diffusion Models
Generative Models
ML Interatomic Potentials
LLM & Agentic Systems
RAG (Retrieval-Augmented Generation)
SFT & RLHF Fine-tuning
Agentic Pipelines
Prompt Engineering
LangChain
MLOps & Engineering
End-to-End ML Pipelines
A/B Testing
REST APIs
Cloud HPC
Python
C++
Scientific Computing
DFT (Density Functional Theory)
Molecular Dynamics
Microkinetic Modeling
Uncertainty Quantification

Languages

Chinese
Native speaker
English
Professional