Ethan Boneh
EE & CS @ Stanford
About
Hi! I'm Ethan, an Electrical Engineering and Computer Science student at Stanford. I'm a research assistant at the Scaling Intelligence Lab (SAIL) and previously interned on Apple's Platform Architecture team. I'm interested in AI systems, GPU kernel optimization, and hardware-software co-design. My recent work spans evolutionary search methods, domain-specific languages for GPU programming, and autonomous semiconductor fabrication. Outside of work, I like old movies, magical realism, tennis, and playing surf rock.
Publications
Nathan Paek, Simon Guo, Vishnu Sarukkai, Willy Chan, William Hu, Ethan Boneh, Simran Arora, Ludwig Schmidt, Kayvon Fatahalian, Azalia Mirhoseini
ICLR 2026 Workshop on Test-Time Updates
Shu Liu, Shubham Agarwal, Monishwaran Maheswaran, Mert Cemri, Zhifei Li, Qiuyang Mang, Ashwin Naren, Ethan Boneh, Audrey Cheng, Melissa Z. Pan, Alexander Du, Kurt Keutzer, Alexandros G. Dimakis, Koushik Sen, Matei Zaharia, Ion Stoica
arXiv Preprint
Fabia Farlin Athena, Jimin Kang, Matthias Passlack, Nathaniel Safron, Didem Dede, Koustav Jana, Balreen Saini, Xinxin Wang, Shuhan Liu, Jonathan Hartanto, Ethan Boneh, Hugo J.-Y. Chen, Chi-Hsin Huang, Qing Lin, Donglai Zhong, Kaitlyn Leitherer, Paul C. McIntyre, Gregory Pitner, Iuliana P. Radu, H.-S. Philip Wong
IEEE Transactions on Electron Devices, 2025
Jack C. Evans, Fabia F. Athena, Koustav Jana, Balreen Saini, Shuhan Liu, Ethan Boneh, Paul C. McIntyre, H.-S. Philip Wong
Device Research Conference (DRC), 2025
Projects
AI-driven closed-loop system that autonomously designs, executes, evaluates, and improves semiconductor fabrication experiments. Uses evolutionary search and computer vision to optimize lithography parameters without human intervention.
End-to-end RL pipeline for GPU kernel optimization. Language models generate CUDA kernels, which are evaluated on cloud GPUs via Modal, with results converted into training rewards for distributed LoRA fine-tuning.
Evolutionary optimization for GPU kernels using the Helion DSL. Applies automated search to discover high-performance kernel implementations.
3D graphics engine and driving game on a RISC-V MangoPi board. Built a full rendering pipeline with LookAt projection, occlusion culling, and flat shading, controlled by a real steering wheel via gyroscope.
VR social platform that converts ordinary photos and videos into interactive 3D environments using neural radiance fields and real-time mesh generation. No expensive equipment needed.