Summary
Yiğit Demirağ is a research scientist and PhD candidate at ETH Zürich focused on algorithm–hardware codesign for efficient gradient-based training on neural network accelerators, particularly those using in-memory computing like PCRAM and ReRAM. He combines deep academic research with practical industry experience—now at Google after internships at MILA, Samsung, EPFL and CERN—working on low-bit training, memory-device modeling, and backprop implementations on crossbar arrays. Yiğit has taught a course on spiking neural networks for undergraduates and has collaborated with leading researchers across academia and industry, reflecting a knack for interdisciplinary teams. With 13 years of experience spanning signal processing, SIMD optimization, and neuromorphic hardware, he brings both system-level perspective and low-level optimization skills. A less obvious strength is his history of accelerating scientific computing primitives (e.g., vectorized random number generators) and translating that performance mindset into efficient learning algorithms for novel hardware.
13 years of coding experience
1 year of employment as a software developer
Master’s Degree, Electrical and Electronics Engineering, Master’s Degree, Electrical and Electronics Engineering at Bilkent University
Mugla 75. Year Science High School
Doctor of Philosophy - PhD, Doctor of Philosophy - PhD at Institute of Neuroinformatics, ETH Zurich and University of Zurich
English, Turkish, Italian