Samuel W. Stark - CV

UK/USA dual national

Cambridge, England

Education

October 2022 - Present

Ongoing, planning to submit in March 2026

Secure FPGA Design
Designed a high-throughput cryptography accelerator for checking memory accesses at the edges of the interconnect. Verified the design using a custom C++ UVM testbench with Verilator. Integrated with a pre-existing soft-CPU interconnect design on a Stratix-10 FPGA.
Low-Level Operating System Integration
Wrote drivers and other software for interacting with the cryptography accelerator in multiple operating systems, including FreeBSD, using C, C++, Rust, and Python.
Writing and Communication
Wrote an article for ACM Queue on high-performance memory safety in CXL, which was subsequently published in Communications of the ACM - the official magazine of the world’s largest scientific computing society. The article is available online.

October 2021 - June 2022

Large-Scale Data Processing and Systems Research
Evaluated and presented research papers on many areas in large-scale data processing and cloud computing, focusing on GPU-based data processing. Profiled and replicated results from an open-source CPU/GPU stream processing research project.
RISC-V Vector Emulation & Security
Built a RISC-V CPU emulator with support for the “V” vector extension, the CHERI memory safety extension, and a combination of the two. Added support for combined instructions to Clang/LLVM, tested using C++ inline assembly. The project is available online.

October 2018 - July 2021

First Class - Best overall graduating student in subject

High Performance CPU and GPU Programming
Achieved best-in-class (64x) speedup on a C fluid simulation with bit-for-bit accuracy using multithreading and vectorized assembly code. Moved the fluid simulation to the GPU using CUDA, with Vulkan graphics for real-time visualization, for my final-year project. The project is available online.
High Performance FPGA Design
Built a high-speed video processing filter on an FPGA using SystemVerilog with various parameters controlled in real-time from the onboard hard Arm CPU.
Machine Learning
Learned basics of classic machine learning and neural networks/deep learning.

Summers of 2019 & 2021

Simulated impact of non-volatile memory and DVM operations in gem5 using C++ and the ‘Ruby’ DSL.

Added DVM operations to existing gem5 CHI protocol model
Simulated impact of DVM on many-core systems
Presented results to key stakeholders and senior design members, including a ‘CHI protocol crash course’
Changes upstreamed to public open source gem5 repository (available online)

September 2017 - August 2018

Developed 3D game engine in C++/Vulkan/Metal with two senior engineers.

2020 - Present

Python addon for importing and exporting the proprietary GMD file format, used by the ‘Yakuza’ game series, into Blender.

Reverse-engineered parts of the file format with the community
Designed algorithms to convert computer-friendly meshes to artist-friendly meshes
Developed beginner-friendly user experience for new modders