Mark Shields
Software Engineer at Modular
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolMark Shields is a software engineer based in Seattle specializing in compilers, tensor dataflow languages, and deep learning runtimes, with a career spanning systems work at Google, ML compiler leadership at OctoML, and current work on DL runtime/compiler intersection at Modular. He led TVM’s front-end CORE team, reworked Relay lowering and device planning, and started the Collage project to enable flexible partitioning and global tuning across libraries and toolchains. His strengths are low-level performance engineering, memory and GPU management, and designing incremental lowering pipelines that bridge functional IRs and imperative tensor codegen. At Google he built large production systems for routing and telemetry, giving him a rare combination of ML/compiler research depth and production service experience. Mark’s contributions to the widely used Apache TVM project include interpreter refactors and debugger tooling that materially improved performance and observability. He pairs an academic pedigree in computer science with pragmatic C++ and Python implementation chops across both research and production environments.
4 years of coding experience
20 years of employment as a software developer
PhD, Computer Science, PhD, Computer Science at Oregon Health & Science UniversityPostdoctoral researcher, Computer Science, Postdoctoral researcher, Computer Science at Microsoft Research, CambridgeThe University of MelbourneBachelor of Science, Computer Science, Bachelor of Science, Computer Science at Monash University
Github Skills (13)
performance-tuning10
performance-monitor10
performance-analytics10
performance-analysis10
performance-monitoring10
performance-measurement10
cprogramming-language9
machine-learning9
c-language9
deep-learning9
compiler-compiler8
compiler8
gpu7
Programming languages (1)
Github contributions (5)
apache/tvm
Jul 2021 - Jul 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:350 reviews, 58 commits, 47 PRs in 1 year
Contributions summary:Mark primarily focused on optimizing the Relay interpreter and compiler within the TVM project. Their contributions included refactoring the interpreter to treat lowering as an IRModule rewrite and implementing a VLOG system for more fine-grained control of debugging information. They also made memory management improvements to support memory-intensive tasks and address performance regressions related to constant handling, especially in relation to GPU memory management.
metalvulkancompilertensoropencl
mbs-octoml/mbs-tvm
Feb 2022 - Jul 2022
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:8 commits, 2 PRs, 619 pushes in 5 months
cpugpu-accelerationtvmdeep-learninggpu
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.