Ganesh Ajjanagadde
About
CV
Contact
Research
Publications
Other
Hobbies

Ganesh Ajjanagadde

photo 

I am a software engineer working in AI SW-HW codesign at Meta. Broadly, I work on

  • performance optimization for Meta's proprietary AI accelerator MTIA.

  • codesign novel ML models/algorithms to make efficient use of HW (e.g., 2nd order optimizers such as Shampoo)

  • performance optimization for existing training workloads (mostly ranking/recommendation) on GPU

  • performance modeling/projection for future GPU, MTIA above

Prior to this, I was a power and performance engineer at Apple working on Siri, responsible for making Siri run fast while not consuming too much power. I am especially interested in efficient algorithms that make great use of available hardware.

Before joining Apple, I was a security/privacy engineer at Snap Inc.. At Snap, I work on end to end encryption, privacy preserving ads solutions, and other problems at the intersection of cryptography and software engineering at scale.

Before joining industry, I defended my Ph.D. thesis titled “Fourier Analysis on the Hypercube, the Coefficient Problem, and Applications” on April 15, 2020.

My graduate work in the Electrical Engineering and Computer Science (EECS) department at MIT was supervised jointly by Prof. Gregory Wornell in the Signals, Information, and Algorithms Laboratory within the Research Laboratory of Electronics (RLE), and Prof. Henry Cohn, a senior principal researcher at Microsoft Research, New England and an adjunct professor in the MIT Department of Mathematics.

Prior to this, I completed my undergraduate work in the EECS 6-2 program at MIT. While an undergraduate, I worked with Prof. Alan Willsky during my freshman year. I also worked with Prof. Yury Polyanskiy, with whom I completed a 'Super’ UROP.

My research during graduate school spanned a wide range of mathematical problems, grouped nicely under the heading point configurations. I thought about such questions in discrete spaces, especially Hamming space. Naturally, techniques for approaching such questions are varied, ranging over probability, combinatorics, number theory, and numerical computation.