Diego Caballero
Principal Deep Learning Compiler Engineer at NVIDIA
San Diego, California, United States
Summary
🤩
Rockstar🎓
Top SchoolDiego Caballero is a Principal Deep Learning Compiler Engineer with 8+ years of experience specializing in vector/SIMD architectures, parallel programming, compilers, and runtime systems. He holds a Ph.D. focused on enabling exploitation of vector instructions in OpenMP and has driven ML compiler work across industry leaders including NVIDIA, Google, and Intel. Diego’s contributions span MLIR, LLVM, IREE and PlaidML, with practical wins such as AVX2 transpose lowerings, cache-level tiling, and Stripe-to-Affine conversions that improve code generation and memory access efficiency. Based in San Diego, he combines deep academic roots from UPC with production-focused engineering on x86, Arm and RISC-V CPU targets, and a knack for turning architecture-aware research into deployable compiler optimizations.
8 years of coding experience
13 years of employment as a software developer
UPC Universitat Politècnica de Catalunya5-year Degree in Computer Engineering (B.Sc. + M.Sc. equivalent), 5-year Degree in Computer Engineering (B.Sc. + M.Sc. equivalent) at Universidad de Murcia
English, Spanish, Catalan
Github Skills (16)
avx10
compiler10
code-generation10
compiler-compiler10
c-language10
cprogramming-language10
vectorization10
mlr10
llvm9
performance-optimization9
machine-learning9
risc-v9
deep-learning9
arm9
type-conversion9
Programming languages (7)
Github contributions (5)
NervanaSystems/ngraph
Feb 2019 - Jul 2020
nGraph - open source C++ library, compiler and runtime for Deep Learning
Role in this project:
Back-end Developer
Back-end DeveloperContributions:113 commits, 85 PRs, 203 pushes in 1 year 4 months
Contributions summary:Diego's primary contribution appears to be the introduction of the `CPURuntimeContextCG` class and associated code generation functionality. This involved modifying `cpu_external_function.cpp` and `cpu_call_frame.cpp` to remove dependencies between nGraph and generated code in code generation mode. They also removed dead code and addressed issues in existing code related to MKLDNN kernels. The user's work focused on optimizing the nGraph library for deep learning.
inference-enginecppc-librarydeep-learningtvm
iree-org/iree
Apr 2022 - Jan 2023
A retargetable MLIR-based machine learning compiler and runtime toolkit.
Role in this project:Back-end Developer & Compiler Engineer
Back-end Developer & Compiler EngineerContributions:458 reviews, 28 commits, 205 PRs in 10 months
Contributions summary:Diego's contributions primarily revolve around enhancing the IREE compiler, focusing on optimization and code generation. Their work includes enabling and improving the AVX2 lowering for transpose operations, addressing issues related to shape casting and code generation driver stages, and adding tests and benchmarks for AVX2 targeting. Further contributions involve setting up a cache level tiling to improve the performance of existing code.
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.