Arya Tschand

Member of Technical Staff at Modal.

Hi! I'm Arya, a Member of Technical Staff at Modal, on the inference research team.

Previously, I was a second-year Computer Science PhD student at Harvard advised by Vijay Janapa Reddi and supported by the NSF GRFP. My research focused on co-designing emerging computer architectures, system architectures, and ML model architectures to unlock orders-of-magnitude gains in inference energy efficiency.

During my PhD, I also held multiple industry research internships.

• Nvidia (Architecture Research Group) - Rubin + LPU inference disaggregation for subquadratic attention LLMs
• Google Research - Autonomous TPU kernel generation (JAXBench) and internal Gemini kernel optimization tools
• AMD (Research & Advanced Development) - Autonomous GPU memory locality kernel optimizations

Before my PhD, I graduated with distinction from Duke with a double major in Electrical & Computer Engineering and Computer Science. I was advised by Dan Sorin and worked on statistically rigorous evaluation methods for computer architecture performance and security.

Feel free to contact me at arya@modal.com.