Mausoom Sarkar

AI Researcher

Noida, Uttar Pradesh, India19 yrs 9 mos experience

Most Likely To SwitchHighly Stable

Key Highlights

15 years of expertise in Deep Learning applications.
40+ patents and publications in top conferences.
Specialized in performance optimization for ML systems.

Stackforce AI infers this person is a Deep Learning and Computer Vision expert in the SaaS and MarTech industries.

Contact

Skills

Core Skills

Machine LearningDeep LearningComputer VisionPerformance Optimization

Other Skills

Large Language Models (LLM)Document Image ParsingAutomated Form ConversionGPU OptimizationCPU OptimizationAlgorithmsImage ProcessingData StructuresComputer ScienceCJavaJavaScriptOpenCLOpenCVCUDA

About

I bring fifteen years of deep expertise in Deep Learning and its real-world applications. My work has resulted in 40+ patents and publications at leading conferences including CVPR, ECCV, ICLR, KDD, ICCV, AAMAS, and WACV.Beyond research, I specialize in performance optimization—from SIMD implementations on CPU (SSE, AVX) to GPU acceleration using OpenCL and CUDA. This dual focus allows me to bridge cutting-edge research with production-ready systems.

Experience

19 yrs 9 mos

Total Experience

6 yrs 7 mos

Average Tenure

17 yrs 9 mos

Current Experience

Adobe

3 roles

Principal Scientist

Jul 2021 – Present · 4 yrs 10 mos

Large Language Models (LLM)Machine LearningDeep Learning

Senior Computer Scientist 2

Jan 2019 – Jul 2021 · 2 yrs 6 mos

Leading the Computer Vision Research under Digital Marketing BU in Adobe.
Created a deep neural network architecture which can parse a document image and can extract the structure present in the document. It is capable of extracting out structure at different granularities and can figure out elements composed of finer level elements up to 3 levels deep.
This research is shipping part of automated form conversion service
https://docs.adobe.com/content/help/en/aem-forms-automated-conversion-service/table-of-contents/introduction.html

Deep LearningComputer VisionDocument Image ParsingAutomated Form Conversion

Senior Computer Scientist

Aug 2008 – Jan 2019 · 10 yrs 5 mos

Machine Learning,Deep Learning
Doing research on computer vision and its applications for Digital Marketing.
Created a deep learning method for automatic creation of shoppable media.
https://blogs.adobe.com/digitalmarketing/web-experience/adobe-transforming-shoppable-experiences/
This method can also be used to do product image search from a single photograph.
Co-Creator of a novel learning to learn technique using Introspection Network.(https://openreview.net/pdf?id=Hkg8bDqee)
GPU & CPU optimizations.
Accelerated multiple Photoshop features using GPU & CPU optimizations.
Implemented the first OpenCl accelerated Photoshop feature the Blurb gallery
Optimized GPU code of Caffe's PRelu layer