Matthias Springer
Deep Learning Compiler Engineer at NVIDIA
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolMatthias Springer is a Deep Learning Compiler Engineer based in Zurich with 13 years of systems and compiler experience, now working on CUDA Tile IR at NVIDIA. He has a strong MLIR/LLVM pedigree from Google where he contributed over 1,000 commits to MLIR-based compilers like IREE, specializing in bufferization, loop transformations, and SSA value bounds inference. Matthias blends deep research credentials (PhD-level training in mathematical and computing sciences) with practical backend engineering, routinely adapting complex codebases to upstream API changes in projects such as Triton and LLVM. His open-source work shows a focus on making compiler infrastructures more robust and performant—adding op support, fixing subtle API regressions, and improving memory and alignment optimizations. He brings low-level CUDA/LLVM expertise together with a history of TPU and XLA performance tuning from earlier Google research and internships. Pragmatic and detail-oriented, he often surfaces non-obvious correctness and performance fixes in large, evolving compiler stacks.
13 years of coding experience
6 years of employment as a software developer
High School Diploma, High School Diploma at Josef-Hofmiller-GymnasiumUniversity of California, San DiegoDoctor of Philosophy - PhD Mathematical and Computing Sciences, Doctor of Philosophy - PhD Mathematical and Computing Sciences at Tokyo Institute of TechnologyMaster of Science (MS) IT Systems Engineering, Master of Science (MS) IT Systems Engineering at Hasso Plattner InstituteDesign Thinking / Stanford Design Innovation Process, Design Thinking / Stanford Design Innovation Process at Stanford University
English, German, Japanese
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (19)
c-language10
operation10
llvm10
tensorrt10
mlr10
intermediate10
triton10
compiler-design10
multilevel10
tensorflow10
compiler-optimization10
tensor10
cprogramming-language10
type-system9
python8
Programming languages (10)
Github contributions (5)
llvm/llvm-project
Jan 2023 - Jan 2023
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:783 reviews, 5 commits, 816 PRs in 6 days
Contributions summary:Matthias's contributions center around enhancing the LLVM compiler infrastructure, particularly within the MLIR (Multi-Level Intermediate Representation) project. Their primary focus is on extending and improving the `ValueBoundsOpInterface`, which involves adding support for new operations, such as `arith.select`, simplifying existing implementations for operations like `scf.for` and resolving issues related to nested definitions and memory allocation. They have demonstrated proficiency in extending the capabilities of existing types and improving the overall robustness and performance of the compiler.
compilerstechnologiesclangsubmittoolchain
iree-org/iree
Nov 2021 - Dec 2022
A retargetable MLIR-based machine learning compiler and runtime toolkit.
Role in this project:Back-end Developer
Back-end DeveloperContributions:152 reviews, 89 commits, 209 PRs in 1 year 1 month
Contributions summary:Matthias implemented and refined bufferization mechanisms within the IREE compiler, specifically focusing on optimizing tensor operations for memory efficiency. This work involved implementing op interfaces for loading and storing tensor data, as well as integrating tensor initialization elimination to reduce memory overhead. The user also contributed to optimizations, such as generating memref.assert_alignment operations to improve data alignment and bufferization of linalg_ext.fft and reverse operations.
mlirspirvvulkantensorflowcompiler
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.