2013 – 2016 · Astrophysics & cosmology
Illustris · dwarf galaxies
UC Riverside · Graduate Research Assistant
Entry into large-scale numerical simulation. Developed a theory for the formation of dwarf galaxies and analyzed terabytes of cosmological data from the Illustris Simulation Suite — one of the largest cosmological N-body + hydrodynamics simulations ever run. The numerical-methods and HPC training that defined the rest of my career started here. First peer-reviewed publication in MNRAS. The experience set the lifelong question: how do you reproduce, in other domains, what Illustris achieved for the universe?
- Astrophysics & cosmology
- Numerical methods & HPC
2016 · Astrophysics & cosmology
MNRAS · dwarf galaxies & globular clusters
Monthly Notices of the Royal Astronomical Society · Lead author
Lead-author MNRAS paper on the assembly of dwarf galaxies in clusters and the efficiency of their globular-cluster formation. The work predicted an excess of globular cluster formation in active, starburst-driven environments. That prediction has since been observationally confirmed multiple times by NASA and other observatories. The paper closed the astrophysics chapter of my career and pointed the way to the methods-driven simulation work that followed at UCSB.
2016 – 2020 · Numerical methods & HPC
PhD · UCSB · level-set PDE methods
UC Santa Barbara · Graduate Research Assistant · CASL Lab
PhD in Mechanical Engineering / Computational Science & Engineering. Authored two state-of-the-art parallel HPC simulation codebases in C++/MPI — one for cell-aggregate electroporation, one for epitaxial growth — both published in the Journal of Computational Physics. Developed a reduced-order stochastic theory for interfacial polarization of cell aggregates. Specialized in level-set methods for irregular free-boundary elliptic PDEs — the same machinery that maps directly onto epitaxy, etching, and lithography in semiconductor manufacturing.
- Numerical methods & HPC
- Biophysics & life sciences
2017 – 2018 · Numerical methods & HPC
Epitaxy · island dynamics on quadtree grids
UCSB · CASL Lab · Lead author
Lead-author paper "The island dynamics model on parallel quadtree grids" (Journal of Computational Physics, 2018). A state-of-the-art parallel C++/MPI simulator for epitaxial thin-film growth using the island-dynamics model, second-order level-set on non-graded adaptive Cartesian grids, with Robin boundary conditions for asymmetric Ehrlich–Schwoebel barriers. The mathematical machinery is canonical to semiconductor manufacturing — epitaxy, etching, lithography — and is the direct lineage to the chip-design work I do today.
2017 · AI for simulation
Physics-aware NN — first investigation
UCSB · CASL Lab · Researcher
Started investigating physics-aware neural networks during my PhD. Explored embedding numerical PDE machinery directly into NN layers, with hybrid optimization for surrogate modeling and early neural-operator ideas. The research line that culminated in BiPDE (Blended inverse-PDE, JCP 2021) and JAX-DIPS (JCP 2022) — and that I still ride at NVIDIA Modulus and on top of the agentic EM simulation platform I build today.
- AI for simulation
- Numerical methods & HPC
2018 – 2019 · Numerical methods & HPC
AVICENNA · tissue-scale simulation
UCSB · CASL Lab · Co-investigator
Architected AVICENNA — the first parallel, multi-scale, multi-physics 3D tissue simulation engine. Solved tissue-scale electroporation on national supercomputers, modeling ion concentrations, electro-permeabilization, and cell-aggregate dynamics. Published in the Journal of Computational Physics, 2018 (parallel Voronoi-based mesoscale simulation of cell-aggregate electropermeabilization). Featured by the U.S. Army Research Lab, TACC, and XSEDE; press coverage in HPCwire, EurekAlert, Phys.org, Science Daily, and Futurity.
- Numerical methods & HPC
- Biophysics & life sciences
2021 · AI for simulation
BiPDE · physics-aware neural networks
Journal of Computational Physics · Co-author
Co-authored BiPDE (Blended inverse-PDE) — a framework that places a hard-coded PDE solver as a custom layer inside a semantic-autoencoder neural network, in contrast to standard PINNs that only add the PDE to the loss. Demonstrated recovery of variable diffusion coefficients in Poisson problems (1D, 2D) and the time-dependent nonlinear Burgers' equation, robust to noise. Published in the Journal of Computational Physics, 2021 (arXiv preprint, January 2020).
- AI for simulation
- Numerical methods & HPC
2022 · Numerical methods & HPC
JAX-DIPS · differentiable PDE solver
Open source · lead developer · Creator
Built and open-sourced JAX-DIPS — a differentiable 3D interfacial PDE solver in JAX that runs on GPU/TPU/CPU. Trains compact neural networks to solve elliptic PDEs with jump conditions, ubiquitous in life sciences and materials. The Neural Bootstrapping Method (NBM) for finite discretization is at the core of the method. Published in 2022 (Journal of Computational Physics, with the related Neuro-symbolic PDE solver appearing at NeurIPS 2022 ML4PS).
- Numerical methods & HPC
- AI for simulation
2020 – 2021 · Biophysics & life sciences
Postdoc · biotherapeutics HPC
Merck Research Laboratories · Postdoctoral Research Fellow
Postdoctoral fellow at Merck. Developed a CUDA/C++ Monte Carlo code using Mayer's sampling for the second virial coefficient of biologics, and proposed a continuum mathematical model for protein aggregation at high concentrations. Awarded a $75,000 seed grant by Merck Research Laboratories leadership to build an HPC platform for protein aggregation — the foundation for industrial-scale biotherapeutic formulation simulation.
- Biophysics & life sciences
- Numerical methods & HPC
2021 – 2023 · AI for simulation
NVIDIA · Clara, Modulus, Warp
NVIDIA · AI & HPC · Senior Software Engineer
Founding member of the Clara Discovery Simulation engineering team for AI-driven drug discovery. Implemented DiffDock — score-based diffusion with 3D-equivariant GNNs — inside the BioNeMo LLM framework. Developed GPU-optimized frameworks for ML-driven molecular dynamics, achieving 5× speed-up and 10× memory efficiency. Conceptualized and led AI-accelerated biophysical simulations for macromolecule stability and solvation free energy. Worked cross-functionally across NVIDIA Modulus (physics-informed ML), Warp (GPU simulation library), and Clara — operating directly at the simulation × AI convergence.
- AI for simulation
- Numerical methods & HPC
- Biophysics & life sciences
Jul – Oct 2023 · Biophysics & life sciences
Prescient Design · protein design
Genentech · ML Scientist
Short residency with the Prescient Design group on therapeutic protein design. The experience clarified where AI-for-drug-design was ready — and where it wasn't — and started me thinking about which domain to bring large-scale physics simulation and AI to next. That domain turned out to be chip design.
- Biophysics & life sciences
- AI for simulation
Jan – Apr 2024 · AI for simulation
Multi-objective DPO for protein design
Aikium · Principal Deep Learning Scientist
Founding member of Aikium's computational team. In four months I invented a multi-objective Direct Preference Optimization (DPO) algorithm for protein and molecule language models and filed three provisional patents on it. After I left, additional team members extended the work, and roughly two years later the algorithm became the foundation of Aikium's production drug-design platform. Published an ICLR 2024 GEM workshop paper on the method; also built a RAG pipeline for protein-design knowledge retrieval.
- AI for simulation
- Biophysics & life sciences
2024 – 2025 · Chip design & semiconductors
Synopsys · founding engineer, GenAI Solutions
Synopsys · Senior Staff AI Engineer
Founding engineer of the GenAI Solutions team at Synopsys. Architected and shipped an internal GenAI SDK (RAG + Agents) used across multiple business units; pioneered a novel multi-agent framework for EDA workflows. Technical lead for a cross-functional initiative building physical-design circuit foundation models — invented a circuit tokenization scheme inspired by SMILES representation of molecules. Developed an AI recommender system for EDA simulator parameter tuning, combining reinforcement learning, transformers, and hyper-graph neural networks. Lead author of Synopsys' patent on its multi-agent EDA framework. Recipient of the Engineering Excellence Group Individual Award — selected among ~2,000 engineers — for advancing corporate-wide GenAI strategy.
- Chip design & semiconductors
- AI for simulation
- Numerical methods & HPC
2025 – now · Chip design & semiconductors
Stealth · agentic physics intelligence
Sequoia-backed chip-design startup (stealth) · Machine Learning Researcher
Founding engineer of the physical-design and verification team at a Sequoia-backed chip-design startup (in stealth). Defined, architected, and built an agentic physics-intelligence platform capable of reasoning about physical design. Pairs a multi-GPU electromagnetics simulation engine with an agent scaffold that runs experiments in silico, inspects numerical outputs, and refines hypotheses against them. Also designed a neural operator architecture suited to this class of simulations, and built an RL-with-verifiable-rewards framework as a tool for LLM post-training.
- Chip design & semiconductors
- Numerical methods & HPC
- AI for simulation
2026 · AI for simulation
Stencil · community journal + agent plugin
stencil.pub · Founder & editor
Founded stencil.pub — a community-owned open journal of physics simulation with a companion agent plugin for Claude Code and Codex. The premise: agents are now responsible for most of what gets built, and they should be able to run physical simulations too. Publications are only worthwhile to the extent that they raise the capability of LLM models and coding agents — every essay published carries the skills, atoms, and verifiable reports its author chose to share, and those accumulate into one community-owned, versioned, attributed plugin. One of my central beliefs is that scientific publishing has to be re-grounded around this, and I want to lead the way in shifting publishing practice to match the real needs of our time.
- AI for simulation
- Numerical methods & HPC
