Kaustubh Raste
Consultant at GlobalLogic
Pune, Maharashtra, India
Summary
🤩
Rockstar🎓
Top SchoolKaustubh Raste is a systems-focused software engineer and consultant with over 20 years of experience and 11 years in professional roles concentrating on system software, AI frameworks, and embedded performance work. Based in Pune, he has advanced through senior and staff engineering roles at Imagination Technologies and now consults at GlobalLogic, bringing deep expertise in low-level optimization for MIPS and ARM. His open-source contributions to high-profile projects like OpenBLAS and libvpx demonstrate hands-on speedups—implementing MSA SIMD kernels and fixing bit-exact IDCT/noise issues that materially improve codec and linear-algebra performance. Comfortable across firmware, video codecs, and numerical libraries, he pairs rigorous engineering with a pragmatic focus on measurable performance gains. Unusually for a consultant, he blends academic training in electronics with two decades of production-grade SIMD and architecture-specific tuning.
11 years of coding experience
18 years of employment as a software developer
Master of Engineering (M.Eng.), Electronics, Master of Engineering (M.Eng.), Electronics at Walchand College of EngineeringB.E., electronics, B.E., electronics at Vishwakarma Institute of TechnologyVidyapeeth High SchoolMaster of Engineering - MEng, Electrical and Electronics Engineering, Master of Engineering - MEng, Electrical and Electronics Engineering at Walchand College of Engineering(A Govt. Aided Autonomous Institute),SANGLI-M.S
Github Skills (14)
video-processing10
arm10
assembly10
video-encoding10
blas10
architectures10
mips10
assembler10
computer-architecture10
architecture10
lapack9
linear-algebra9
c119
c179
Programming languages (2)
Github contributions (5)
OpenMathLib/OpenBLAS
May 2016 - Jan 2017
OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:13 commits, 11 PRs, 2 comments in 8 months
Contributions summary:Kaustubh's commits focused on optimizing double-precision floating-point matrix multiplication (DTRSM) and single-precision floating-point matrix multiplication (STRSM) kernels for MIPS P5600 and I6400 architectures. They implemented these optimizations using MSA (MIPS SIMD Architecture) instructions, improving the performance of the BLAS library. These contributions also included fixing bugs related to these optimized functions.
lapackopenblaslapackelinear-algebrabsd
webmproject/libvpx
Oct 2016 - Dec 2017
Mirror only. Please do not send pull requests.
Role in this project:Back-end Developer
Back-end DeveloperContributions:45 commits in 1 year 2 months
Contributions summary:Kaustubh primarily focused on optimizing and fixing potential bit-mismatches within the `webmproject/libvpx` repository. The contributions involve modifying and correcting several functions related to adding noise to planes and inverse discrete cosine transform (IDCT) functionality, specifically for MIPS and ARM architectures. The changes involved edits to multiple files within the `vp8` and `vp9` directory structures, suggesting a focus on core video encoding/decoding routines. These fixes aimed to improve the accuracy and performance of specific image processing steps during video encoding and decoding.
pull-requests
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.