AI Research Scientist @ Axiomatic AI · Duke PhD in Physics (ECE)

I build AI systems that replace expensive physical simulations with fast, accurate surrogate models — from metamaterials to nanophotonics.

🔬 ML_MM_Benchmark — Open-source benchmarking suite for DL surrogate simulators across 3 physics problems. MLP, Transformer, and MLP-Mixer with pre-trained models. Accepted at NeurIPS 2021 Datasets & Benchmarks Track. 16 ⭐ · 7 forks · pip install AEML

📄 Physics-Informed Learning in Artificial Electromagnetic Materials — Applied Physics Reviews (2024). Survey of ML techniques (Bayesian optimization, inverse design, active learning) that accelerate metamaterial design by 5 orders of magnitude.

📄 Can Large Language Models Learn the Physics of Metamaterials? — IEEE Access (2024). Empirical study on foundation models for high-dimensional scientific regression.

• 💡 AI research for photonics design and simulation at Axiomatic AI
• 🔧 Agentic AI systems for physics simulation
• 🚀 Building toward shipping a side project by late 2026

• PhD — Duke University, ECE. Thesis: Machine Learning for Next Generation Metamaterials. Advisor: Willie J. Padilla. 10+ publications including Nature Sustainability, Applied Physics Reviews, NeurIPS.
• BS — University of Rochester, Optical Engineering.
• Industry — Axiomatic AI (AI research scientist, photonics) · KronosAI (MTS) · Metalenz (computational EM & photonics).
• Awards — Best Student Talk (Duke ECE), Energy Data Analytics PhD Fellow, Best Grad Poster (Triangle Hard Matter Workshop).

Languages: Python · Julia · C/C++

ML / DL: PyTorch · TensorFlow · scikit-learn · NumPy · pandas

Physics Simulation: RCWA · FDTD · FDFD · CST · COMSOL · CodeV/Zemax

Infra: Slurm · AWS · Git · Linux · Docker

AI-Augmented Dev: Claude Code · Cursor · OpenAI Codex · MCP

Agent Systems: Agentic pipelines · Sub-agents · Automated eval harnesses

_{This README was crafted with assistance from GPT-5.4 and Claude Opus 4.6 — because why not use the tools you build with.}