Jul 2019 – Apr 2023 · 3 yrs 9 mos

• Worked on design and development of Neural Network Library in C++. It includes Feed forward and back propagation for Dense layer and convolution layer.
• Improvements in Deep Learning Models for Action detection using unsupervised learning.
• Worked on Run time optimization, Research and development of new algorithms in computer vision/ Image Processing:
• Design and implement a new algorithm for detection of Blockiness in video.
• Design and implement a new algorithm for detection of Ghosting Artifact in video.
• Run time performance enhancement of product named Baton.

Jan 2017 – Jun 2019 · 2 yrs 5 mos

• Worked on development of new techniques and methods in computer vision and machine learning in Content Classification tool, Winnow. My responsibilities includes:
• Action detection(Fight Action) in video to classify content as violence or no-violence using Deep learning and computer vision techniques.
• Gunshot detection in video using Deep learning and computer vision techniques.
• Worked on Architecture design , Run time optimization, Research and development of new algorithms in computer vision/ Image Processing.
• Design and implement a new algorithm for detection of Stuck Blocks dropouts in video.
• Design and implement a new algorithm for detection of Defective Pixels / Hot Pixels/ Dead Pixels in video
• Design and implement a new algorithm for dropout errors due to erroneous concealment by decoder

Jan 2015 – Jan 2017 · 2 yrs

• Research and development of automated checks for no reference objective video quality assessment in QC tool, Baton. This included designing algorithms to process uncompressed videos and detection of various artifacts and errors in a video/image.
• Enhancement in existing algorithm for detection of Black Frame in a video
• Enhancement in existing algorithm for detection of Freeze Frame in video
• Enhancement in existing algorithm for detection of Shot Transition in video
• Enhancement in algorithm for detection of Blocky Dropouts in video to remove False Detections
• which includes Enhancement in motion detection techniques.
• Enhancement in algorithm for detection of Chroma Dropout in video for both improvements in
• detection and to remove False Detections.
• Implementation/Designing a new algorithm for detection of Vertical Line Patterns Dropouts in
• video
• Implementation/Designing a new algorithm for detection of Blank Frames in video