Aug 2016 – Jul 2022 · 5 yrs 11 mos · Bangalore

• Working on C, (PP) Pixel Processor (Broadcom), Broadcom specific assembly language, Optimization of image Processing Algorithms like : Pyramids, HOG, SVM, Feature Concatenation etc
• Projects I worked on:
• A - Project – Pedestrian Detection (PED)
• Client: Daimler Duration: Jan 2017 - till now
• Role: Embedded Algorithms & Optimization Lead
• Rear view pedestrian detection during parking, Assisting driver to apply break.
• Skilled Used:
• C, Pixel Processor(PP), Broadcom assembly language, PP optimisation.

Aug 2014 – Aug 2016 · 2 yrs · Bangalore

• Working on C, C++, C++AMP, NEON Optimization of image Processing routines.
• Projects I worked on:
•  Project – JPEG Optimization and Gallery PLM Support:
• Client: Samsung Duration: Jan 2015 - Aug 2016 Role: Lead Engineer
• JPEG an image Compression technique in Gallery, Which is a lossy compression method for color image.
• Skilled Used:
• C, C++, C++AMP, Data Structure, NEON, JPEG, Image Post Processing and Multi-Threading
• Achievements:
• Ground up basic JPEG Compression method.
• C, C++AMP and NEON Intrinsic Optimization, which result in 25-30% improvement in performance of Huffman Decoding Module.
• Successfully complied with Agile process.
•  Project – HDR- High Dynamic Range :
• Client: Samsung Duration: Aug 2015 - Dec 2015 Role: Lead Engineer
• Providing the HDR as an option to Camera mode, which enable the facility to capture Dynamic Range of scene
• Skilled Used:
• C, C++, Data Structure, Android Native, Image Processing, C++ AMP and NEON
• Achievements:
• Ground up basic development and program flows.
• Android Application development.
• C, C++AMP and NEON Intrinsic Optimization, which result in 30-40% improvement in performance.
• Benchmarking and Documentation of project.
•  Project – 3D Scanning :
• Client: Samsung Duration: Oct 2014 - Aug 2015 Role: Lead Engineer
• 3D reconstruction of an object using Mobile Camera.
• Skilled Used:
• C, C++, Data Structure, Neon Optimization and image understanding
• Achievements:
• Ground up basic development and program flows.
• C++AMP and NEON Intrinsic Optimization of texture mapping module, which result in 20-30% improvement in performance.
• Android NDK coding and porting
• Benchmarking and Documentation of project.

Jul 2012 – Aug 2014 · 2 yrs 1 mo · Bangalure, Karnataka, india

•  Project – Math Library(LibM) :
• Client: AMD India Pvt. Ltd. Duration: Jul 2012 - Jan 2013 Role: Software Engineer-2
• Collection of basic math functions like Exp, Log and Trig group of functions (Scalar and Vector Variant). http://developer.amd.com/tools-and-sdks/cpu-development/libm
• Skilled Used:
• C, Linux, Perl, Computer Architecture(AMD), X86(AVX/FMA3/FMA4/XOP)
• Achievements:
• Improve the routines which result in 30-40% improvement in performance.
• Porting to Linux.
•  Project – AMD Core Math Library(ACML):
• Client: AMD India Pvt. Ltd. Duration Jul 2012 - Jan 2013 Role: Software Engineer-2
• Collection of level 1,2 and 3 BLAS, (LAPACK) routines. Suite of FFTs, Random Number Generators (single&double) http://developer.amd.com/tools-and-sdks/cpu-development/amd-core-math-library-acml/
• Skilled Used:
• C, Linux, Perl, Computer Architecture(AMD), X86 (AVX/FMA3/FMA4/XOP)
• Achievements:
• Porting to Linux
• Wrote Perl Benchmark script to benchmark the routines for different Instruction sets like AVX, XOP, FMA3 and FMA4.
•  Project – BOLT:
• Client: AMD India Pvt. Ltd. Duration Jan 2013 - Jul 2014 Role: Software Engineer-2
• BOLT - OpenCL accelerated C++ Template Library - http://github.com/HSA-Libraries/Bolt
• Skilled Used:
• C, Linux, Python, C++, Data Structure, Design Pattern, OpenCL, TBB, C++ AMP
• Achievements:
• Ground up basic development of routines and program flows.
• Performance improvement of reduce, inner product, scan, min, max and copy routines which result in 40-50% improvement in performance.
• Implemented the OpenCL, C++AMP, TBB and Serial Codes for Scatter, Scatter_if, Gather and Gather_if routines.
• Porting to the Linux.
• Implemented the Command Line Parser to remove the Boost dependency.
• Wrote the benchmark code which benchmark routines for Serial, multi-core, OpenCL, C++AMP and Thrust.
• Wrote Python Script for Performance comparison of the routines which takes different command line arguments.