Ningdong Wang

CV

Download résumé
Ningdong Wang

I’m a physics PhD student at Cornell University, with research in simulation, optimization, and machine learning for particle accelerators. I work across the AI stack, from CUDA and GPU optimization to model training, inference, and agent systems.

My recent work includes models of physical dynamics and tools for automated research and agent evaluation. During my internship at Shizuku AI, I built a real-time multimodal agent harness and optimized text-to-speech inference, reducing time to first audio from 600 ms to 180 ms.

I expect to graduate in 2027 and am looking for full-time AI research and engineering roles. I’m particularly interested in systems that can run experiments, evaluate results, and improve through iteration.

Ningdong Wang · nw285@cornell.edu · ndwang.github.io

Education

2021 — Present

Cornell University

PhD · Physics

Ithaca, NY

Expected May 2027

2016 — 2020

University of Illinois at Urbana-Champaign

B.S. · Electrical Engineering and Engineering Physics

Urbana, IL

Experience

2021 — Present

Cornell University

Research Assistant

Implemented cathode space-charge tracking in Bmad and studied RF cavity coupler effects on beam quality. Develop GPU-accelerated simulation tools in Julia and machine learning surrogates for beam dynamics and electric-field prediction. Develop MCP-connected agents for accelerator simulations and control systems, with evaluations of correctness and reliability on operational tasks. Contribute to Osprey, a multi-institution framework for agentic AI in accelerator control systems. Supervised by Georg Hoffstaetter.

2026

Shizuku AI

Creative Engineer (Summer Intern)

Built a text-to-speech data pipeline that generated 25,000 hours of training audio. Optimized TTS inference, reducing time to first audio from 600 ms to 180 ms. Developed a real-time multimodal agent harness integrating audio, vision, audience comments, tool calls, and speech output.

2023 — 2024

Brookhaven National Laboratory

Research Assistant

Modeled space charge and collective effects for Electron-Ion Collider cooling systems. Optimized the Strong Hadron Cooler injector and merger in Bmad and Xopt, reducing simulated normalized transverse emittance by 62% and RMS energy spread by 85% relative to the original configuration.

2019

Argonne National Laboratory

Lee Teng Intern

Modeled and optimized a spin-rotator solenoid array for the SuperKEKB electron storage ring. Supervised by Uli Wienands.

2018 — 2020

University of Illinois at Urbana-Champaign

Research Assistant

Studied electron spin polarization in quantum point contacts and its effect on conductance. Supervised by Jean-Pierre Leburton.

Recognition

2025

DOE Tigner Traineeship in Accelerator Science

Department of Energy, Office of Science

2023

Skills

Programming

Python · Julia · C · C++ · CUDA · Fortran

Machine learning

PyTorch · Transformers · VAEs · 3D U-Net · Physics-informed neural networks

AI agents

Agent harnesses · MCP · Tool calling · Agent evaluation · RAG

Inference

vLLM · TensorRT · Streaming inference

HPC

Custom GPU kernels · NVIDIA Nsight · MPI · OpenMP · Slurm

Numerical methods

Multi-objective genetic optimization · Bayesian optimization · Numerical analysis · Linear algebra

Systems & tools

FastAPI · Asyncio · WebSockets · Docker · Linux · Git

Accelerator modeling

Bmad · Beam dynamics · Space charge · Particle tracking · Injector and ERL simulation

Publications

2026

2025

2024

2023

2022