Haozhe Zhang

Palo Alto, CA

USA

Hey, welcome to my page! 👋

I’m a modeling engineer at Tesla, working at the intersection of physics-based modeling and machine learning for next-generation battery cells — theoretical and computational modeling, reinforcement learning for automation, and data-driven optimization at scale-up.

My background is heavy on physics. I’ve spent years working out closed-form theoretical solutions across mechanics and multi-physics fields — soft-hard material integration, stretchable metasurfaces, mechanical Janus structures, nanoconfined fluid mechanics. I pair the theory with hands-on engineering, so most of what I’ve built has gone end-to-end from the math to a working system.

Outside the day job, what pulls me from one thing to the next is curiosity and a long-running urge to build new things. I keep gravitating toward LLMs, agentic systems, RL training infrastructure, voice agents for narrow domains, and AI for hardware — all self-taught.

Before Tesla, I earned my PhD in mechanical engineering from the University of Virginia and a BS from the University of Science and Technology of China.

In my spare time, I enjoy traveling and playing Leagues.

news

May 13, 2026	Workshop paper accepted at CVPR 2026 NeXD: “CADLoop: An Equivariant-Aware Skill-Grounded Loop for CAD Data Curation” — a neuro-symbolic pipeline that repairs near-miss CAD programs (IoU ≈ 0.8 regime), turning ~91% Re-Act pass rates into 100% with IoU ≥ 0.99 training pairs.
May 11, 2026	Our new paper “BenchCAD: A Comprehensive, Industry-Standard Benchmark for Programmatic CAD” is now on arXiv. Paper: arXiv:2605.10865 · Writeup: 工业 CAD 远看是 RLVR 的好题目，做起来才发现每一层都很糙.
Mar 15, 2026	New blog: PPO 到 GSPO 究竟在改什么 — RL for LLMs walkthrough including K1/K2/K3 KL estimators and throughput tricks.
Nov 30, 2025	Voice agent for stocks — prototype is up. Writeup: 通勤一小时，做了个聊股票的 voice agent.
Nov 10, 2025	Started building a personal voice agent for stock trading and news — implementation in news_agent.
Aug 15, 2025	Our paper “A buckling mechanics model for pattern transformation of lattice superstructures assembled by soft-hard materials integrated units” has been published in ACS Applied Polymer Materials.
Apr 23, 2025	Our paper “Degradation of Hydrogen Bonds Enormously Enhances Convective Heat Transfer in Nanofluidics” has been published in Nano Letters. Read it here.
Apr 18, 2025	New blog: SVI vs SABR：短期合约 vol smile 上的差距 — comparing two industry-standard volatility models on an AAPL surface.
Sep 20, 2024	Our paper “Extremely durable electrical impedance tomography–based soft and ultrathin wearable e-skin for three-dimensional tactile interfaces” has been published in Science Advances. Read it here.
Feb 10, 2023	Our paper “Mechanical Janus Structures by Soft–Hard Material Integration” has been published featured as inside cover. Here is the link.

featured

BenchCAD: A Comprehensive, Industry-Standard Benchmark for Programmatic CAD

Haozhe Zhang, Kaichen Liu, Miaomiao Chen, Lei Li, Shaojie Yang, Cheng Peng, and Hanjie Chen

2026

Abs URL Project

We present BenchCAD, a comprehensive industry-standard benchmark for evaluating programmatic CAD code generation. The benchmark includes over 17,000 execution-verified CadQuery programs spanning 106 industrial part families (e.g., gears, springs, drills) and assesses models across visual question answering, code analysis, image-to-code conversion, and code editing. Our results show that current models often recover coarse outer geometry but fail to produce faithful parametric CAD programs, with common failure modes including missing structural details and oversimplified operations; generalization to unseen part families remains limited even after fine-tuning.