Felipe Aramburu
Distinguished Solutions Architect at NVIDIA
Houston, Texas, United States
Summary
🤩
Rockstar🎓
Top SchoolFelipe Aramburu is a Distinguished Solutions Architect with nine years of experience building GPU-driven distributed execution runtimes and a track record of shipping systems that operate on data scales far beyond GPU memory (into the 100TB range). As co‑founder and lead architect at Voltron Data and formerly CTO at BlazingDB, he architected asynchronous multi-executor engines, advanced memory allocators, and I/O strategies that delivered orders-of-magnitude performance improvements and scaled to hundreds of GPUs. He combines hands‑on performance engineering—contributing to RAPIDS projects like RMM and cuDF—with product leadership, designing observability and dynamic memory policies that minimize fragmentation and keep GPUs fully utilized. Based in Houston, he brings deep practical expertise in GPU memory/resource integration, RDMA-enabled transfers, and vector search/embedding pipelines, plus a rare ability to prototype allocator and scheduling strategies end-to-end. Notably, his designs have been validated at supercomputer and production cluster scale and influenced open-source GPU data ecosystems.
9 years of coding experience
20 years of employment as a software developer
B.A. Economics B.A. Latin American Studies Computational Economics, B.A. Economics B.A. Latin American Studies Computational Economics at The University of Texas at AustinTaylor HS
Spanish, French, German, Portuguese
Stackoverflow
308reputation
34kreached
9answers
16questions
Github Skills (22)
thrust10
memory-allocation10
c-language10
gpgpu10
memory-management10
rapids10
gpu10
accelerated-computing10
cuda10
cprogramming-language10
data-structure9
algorithms9
data-structures9
data-analysis8
dataframes8
Programming languages (8)
Github contributions (5)
rapidsai/cudf
Jun 2018 - Apr 2019
cuDF - GPU DataFrame Library
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:86 commits, 4 PRs, 2 branches in 10 months
Contributions summary:Felipe primarily focused on implementing and optimizing GPU-based DataFrame operations within the cuDF library. Their work involved adding windowed functions, including the implementation of sorting and hashing operations. They contributed to hash and window operation code, which involved modifications to the build process. They also focused on optimizing performance of various operations by creating optimized data structures and algorithms to utilize the GPUs resources.
cudadataframe-librarydata-analysiscppcudf
rapidsai/rmm
Sep 2019 - Oct 2019
RAPIDS Memory Manager
Role in this project:Back-end Developer
Back-end DeveloperContributions:40 commits, 1 PR, 41 comments in 1 month
Contributions summary:Felipe primarily contributed to the core functionality of the RAPIDS Memory Manager (RMM), focusing on memory resource implementations and allocation strategies. Their work involved integrating various memory resources like CUDA, cnmem, and managed memory, as well as modifying how memory is allocated and deallocated. The user also worked on incorporating features related to memory info retrieval and logging.
cudamemory-managementmemorycpppython
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.