Jiajia Xie

Jiajia Xie

AI, Machine Learning, and Computational Scientist.

I am a machine learning engineer in Energy and Charging’s Reliability Engineering at Tesla, where I build and evaluate LLM-agent systems for reliability engineering and physics-informed machine learning models for prognostics of Tesla’s products.

I obtained my Ph.D. from the interdisciplinary Computational Science and Engineering program at Georgia Tech, advised by Prof. Cassie S. Mitchell in the Laboratory for Pathology Dynamics. I previously earned my M.S. in the same program under the joint supervision of Prof. B. Aditya Prakash and Prof. Lauren N. Steimle, focusing on data-driven modeling and controls of pathology dynamics, epidemics and human behavior. I hold a B.S. in Economics and Mathematics from Presbyterian College (South Carolina, USA). During my Ph.D., I interned at Amazon, Tesla, and Berkeley Lab.

Current Work at Tesla

My work centers on making LLM-agent workflows trustworthy enough to support formal reliability engineering processes.

  • LLM-agent for FMEA. I research and develop AI-agentic workflows that draft failure mode and effects analyses (FMEAs), estabish evaluation layer and benchmark that keeps them aligned with expert judgment and preference, and automate fault tree analysis.

Research Interests

A central theme of my PhD research is bridging modern generative and language models with structured or physical systems under perturbations — from the pathology dynamics of diseases to industrial reliability processes and human mobility. Recently, I have been exploring generative diffusion/flow models, LLM-agent evaluation, stochastic optimal control, and network science.

Bio-mechanistic Generative Models

  • Variational autoencoder–style approaches for generating brain connectomes under neurological diseases (IJMS 2025).
  • Dynamic Brain Connectome Vulnerability in Neurodegeneration via Score-based Network Diffusion (NPJ DEMENTIA).
  • Diffusion Bridge Sampler and Stochastic Optimal Control (title removed for review anonymization).

Additional Work

  • Source-robust non-parametric reconstruction of epidemic-like event-based network diffusion processes with online data (BDCC 2025)
  • Augmenting Bayesian topic models using online confirmations from community-driven apps (BuildSys 2023)
  • Representative deep-gray thermodynamic models of residential buildings (Energy & Buildings 2024)
  • Empirical WiFi datasets for localizing COVID-19 interventions (Frontiers in Digital Health 2023)

Outside of research, I am a dreamer, reader, and enthusiastic audience member of musicals, sports, and concerts (🖤🩷).

Contact