Abhishek Sansanwal
Senior Software Engineer at Microsoft
Noida, Uttar Pradesh, India
Summary
👤
Senior🎓
Top SchoolAbhishek Sansanwal is a Principal Software Engineer with eight years of experience building high-performance multimedia and security systems across industry leaders including NVIDIA, Microsoft, Samsung, and Cadence. He specializes in C/C++ backend development, low-latency audio/video pipelines, and accelerator-targeted kernel work, and has shipped production SDKs for CloudXR and ML-based audio enhancements. At NVIDIA he made notable open-source contributions to DALI’s VideoReader—adding VP9 support, better VFR heuristics, and labeled video handling—which improves video preprocessing for deep learning workloads. At Microsoft he implemented IVP compilation, intercore communication primitives, and a scalable SoC simulator cluster that cut estimated cloud costs by ~$30k/month while enabling ML demos. Based in Noida, he combines systems-level rigor with pragmatic engineering for cross-architecture ports (x64/ARM) and has a track record of turning performance-focused research into deployable tools.
8 years of coding experience
6 years of employment as a software developer
Bachelor of Engineering - BE, Computer Engineering, Bachelor of Engineering - BE, Computer Engineering at Netaji Subhas Institute of Technology
French, English, Hindi
Github Skills (8)
video-processing10
c-language10
cprogramming-language10
gpu10
ffmpeg10
data-processing10
python9
vp99
Programming languages (5)
Github contributions (5)
NVIDIA/DALI
Jun 2019 - Jan 2020
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
Role in this project:
Backend Developer
Backend DeveloperContributions:1 review, 16 commits, 17 PRs in 7 months
Contributions summary:Abhishek primarily contributed to the `VideoReader` operator, adding new features and improving existing ones. They implemented support for reading video files with labels using file lists and enhanced the heuristic for variable frame rate detection. Additionally, they upgraded the VideoCodecSDK and added support for the VP9 codec, improving the overall functionality of the video processing pipeline. Further work included enhancements to label handling and adding options for frame numbers and timestamps.
pythontrainingpaddlegpu-accelerationbuilding-blocks
a-sansanwal/DALI
Jun 2019 - Oct 2020
A library containing both highly optimized building blocks and an execution engine for data pre-processing in deep learning applications
Contributions:88 pushes, 37 branches in 1 year 3 months
pytorchdata-pre-processingpre-processingbuilding-blocksdeep-learning
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.