Seyed [Iman Hosseini] Zavaraki

Software Engineer @ GOOGLE

I started programming GPUs in middle-school & never stopped. Currently making Gemini go brrr on TPUs.

Experience

2024 Sep - Present Software Engineer, GOOGLE, UK
  • ML Compilers: XLA:TPU, Mosaic and Pallas.
  • Tools for low-level profiling of TPU workloads to help development of Pallas kernels and enhance developer experience.
  • Core contributor to Gemini: working with DeepMind's performance and LSP teams to optimize inference and pre-training of Gemini on TPUs. Making the latest TPU features available to kernel developers.
2024 Feb - Sep Compiler Developer, NVIDIA CORPORATION, UK
  • HPC Compilers team. As HPC Compilers' Pilot in Command for Blackwell support, I ensured our compilers (nvc/nvc++/nvfortran) are ready for upcoming GPUs and make best use of latest hardware features. Collaborated with other teams to deliver.
  • Competitive analysis with other compilers.
  • Support for atomics across OpenACC/OpenMP/stdpar: swept up all open bugs and implemented unsupported ones, including necessary support in the frontends.
  • Design, investigation and implementation of novel compiler optimizations to get the best performance on GPUs.
2022 May - 2024 Feb Compiler Developer Intern, NVIDIA CORPORATION, USA
  • Worked on performance of GPU offloading, delivered wins in execution time across the board.
  • Made tools to help with debugging and profiling.
  • Contributed to triaging & fix for internal/externally reported bugs.
2019 - 2024 Lab Member, NYU OSIRIS LAB, USA
  • Presentations and tutorials on hacking and reverse engineering.
  • Authored challenges for CSAW'20 & CSAW'21 CTF.
  • Member of CSAW Applied Research Program Committee.
2019 - 2024 Research Assistant, NYU CENTER FOR CYBERSECURITY, USA
  • Member of MESS Lab, a security lab focusing on Machine Learning, Embedded Systems, and Software/Systems Security.
Oct 2018 - Jan 2019 Software Developer Intern, Central Bank of Iran, Iran
  • This internship counted as part of my credits for the Computer Engineering program.
  • Optimizing statistical & economical calculations in .NET using the Common Language Runtime.
Summer 2018 Software Developer Intern, Vispiron Systems, Germany
  • Worked in the Embedded CoC on infrastructure and algorithms for autonomous vehicles.
  • Designed & implemented a low-latency networking protocol to stream car data (LiDAR+Video).
  • Developed software for a thermal camera mounted inside the car.
Summer 2017 Computer Science Intern, University College Dublin, Ireland
  • Supervised by Dr. Brett Becker. Implementing & assessing models of gravitational waves to estimate frequency range needed to detect BBHs, in collaboration with Dr. John Regan from DCU.

Education

2019-2024 PhD-dropped out, Computer Science, New York University, USA
2022 MS, Computer Science, New York University, USA
2015-2019 BS, Theoretical Physics, Sharif University of Technology, Iran
2014-2019 BS, Computer Engineering, Sharif University of Technology, Iran

Honors & Awards

2022 1st place in SC22 Parallel Programming Marathon in Dallas
2022 2nd place in DEFCON CTF finals in Las Vegas, as member of team Katzebin
2021 Finished 3rd in the A-HUG Hackaton, a week-long cloud Hackaton for Arm-based HPC systems
2019, 2022 Science Alliance Membership at New York Academy of Sciences (NYAS), sponsored by NYU
2019 Awarded NYU School of Engineering PhD Fellowship
2016 Finished 64th in the world (out of 2500 competing teams) in the IEEEXtreme programming contest
2013 Gold Medal in the National Physics Olympiad, Iran

Publications

Beyond the c: Retargetable decompilation using neural machine translation.
Iman Hosseini and Brendan Dolan-Gavitt. NDSS Workshop on Binary Analysis Research (BAR), Hybrid Event, San Diego, CA. 2022