Manupa Karunaratne
Principal Engineer - AI Compilers at Arm
Cambridge, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolManupa Karunaratne is a Principal Engineer specializing in AI compilers with eight years of experience spanning circuit-level design to compiler and runtime optimizations. Based in Cambridge, he has driven ML compilation for ARM and AMD accelerators—contributing upstream to Apache TVM and key MLIR work in IREE and rocMLIR—focusing on attention kernels, vector distribution, and memory-reuse optimizations that boost parallel performance. His PhD work produced novel CGRA architectures and tooling that informed later chip tapeouts and low-power many-core designs, giving him rare fluency across RTL, architecture simulation, and compiler codegen. A proven open-source maintainer and committer, he blends deep research insight with production-grade engineering to solve hardware-software co-design challenges, and often surfaces compiler fixes that enable real-world ML workloads on constrained devices.
8 years of coding experience
7 years of employment as a software developer
Bachelor of Science of Engineering (B.Sc.Eng) Electronics and Telecommunications Engineering, Bachelor of Science of Engineering (B.Sc.Eng) Electronics and Telecommunications Engineering at University of MoratuwaHigh School Advanced Level : Mathematics Physics and Chemistry, High School Advanced Level : Mathematics Physics and Chemistry at D.S.Senanayakee CollegeeDiploma in Management Accounting - CIMA Exams Completed, Diploma in Management Accounting - CIMA Exams Completed at Achievers Lanka Business SchoolDoctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at National University of Singapore
Github Skills (16)
compiler10
machine-learning10
compiler-design10
c-language10
compiler-compiler10
cprogramming-language10
vectorization10
gpu10
mlr10
llvm10
cuda8
tensorflow7
pytorch7
vulkan7
runtimes6
Programming languages (6)
Github contributions (5)
apache/tvm
Mar 2020 - Jun 2022
Open Machine Learning Compiler Framework
Role in this project:
Back-end Developer, ML Engineer
Back-end Developer, ML EngineerContributions:717 reviews, 47 commits, 154 PRs in 2 years 3 months
Contributions summary:Manupa rewrote and modified the Graph Partitioner in the Relay module to support multiple outputs in the deep learning compiler stack. They implemented the necessary changes to support disjoint annotated regions, ensuring functions are created for annotated regions. The user also made codestyle improvements and fixed issues related to tuple outputs in the PartitionGraph, resulting in a more robust deep learning compilation pipeline. They added code and made edits related to workspace allocation in the AoT executor.
compilermachine-learningtensordeep-learninggpu
iree-org/iree
Jun 2024 - Feb 2025
A retargetable MLIR-based machine learning compiler and runtime toolkit.
Role in this project:Back-end Developer
Back-end DeveloperContributions:72 reviews, 37 PRs, 17 pushes in 8 months
Contributions summary:Manupa primarily contributed to the IREE compiler project, focusing on the LLVMGPUVectorDistribute and LinalgExt dialects. Their work involved refactoring existing code, such as the AttentionOpDetail struct, and implementing distribution of vector operations like `vector.step` and `vector.contract`. They also introduced shared memory reuse and improved the distribution of multi-reductions across subgroups. These changes indicate a strong focus on performance optimization and parallelization within the compiler infrastructure.
compilermachine-learningmlirvulkantensorflow
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.