Homepage of David Gens

About

I am a Senior ML Compiler Engineer at Apple.

My interests span a wide range of complex computational systems such as compilers, runtimes, and deep neural networks. I am especially interested in understanding performance characteristics and failure modes of these systems. My past work includes research into self-supervised representation learning, automated compiler-level software analysis, hardware security research, as well as memory-corruption attacks and defenses.

Projects and Publications

The full list of publications is here. I have occasionally served as an external reviewer to conferences and journals (including IEEE S&P, ACM CCS, NDSS, ACM TKDD, and IEEE TDSC), and also given several invited talks. Some recent project examples:

Estimating Stack Frame Sizes using Transformers (ACM CCS'21 - AISec Workshop) in Stripped and Optimized Binaries (AMD64 and AArch64)
Camera Calibration (Comma.ai Challenge 2021) using Convolutional Neural Networks
V0LTpwn (USENIX Security 2020) (remote-fault injection from software on x86 that breaks SGX)
NoJITsu (NDSS 2020 & BH20) (data-only attack against interpreters in JIT engines)
K-Miner (NDSS 2018) (a static analysis framework to uncover memory corruption in Linux at scale using LLVM)

CV

since '26	I now work for Apple, as part of their Neural Engine compiler org!
'22 to '26	Building PyTorch models and MLIR-based compilers for the world's largest AI chip at Cerebras Systems.
'19 to '22	Postdoctoral Researcher at the Secure Systems and Software Laboratory of Professor Michael Franz.
'14 to '19	PhD student at the Intel Collaborative Research Institute and the System Security Lab of Professor Ahmad-Reza Sadeghi.
'13 to '14	Developed trade-workflow and performance-analysis tools at 360T for their OTC/FX-trading platform.
'09 to '13	Studied Mathematics at Gutenberg Universität and Computer Science (B.Sc.) at HSRM.