Feb 2017 – May 2019 · 2 yrs 3 mos · Noida, Uttar Pradesh, India

• Acrobat Desktop: On Device Model:
• 1) Designed and implemented from scratch, an on device ML model interface for
• Acrobat Desktop. This included first integrating Tensorflow APIs in a C++
• environment and then exposing generic and scalable APIs to be used by all
• Acrobat features and plugins. This helped improve the user experience by 80% as
• compared to on server ML
• PDF: Accuracy of CNN model
• 1) Improved the accuracy of CNN based object detection model for PDF documents
• by implementing many new ideas in training data and post processing decision
• trees. Achieved 50% improved accuracy in mAP scores for the model.

Jun 2015 – Jan 2017 · 1 yr 7 mos · Noida, Uttar Pradesh, India

• Acrobat Desktop : Rendering & Performance
• 1) Designed and Implemented from scratch a cloud based, scalable
• infra for performance certification of production builds
• 2) Improved PDF Find feature performance by up-to 25% on user files.
• 3) Implemented Link Time Optimization for CEF (Chromium Embedded Framework)
• dlls and improved PDF rendering time to half of before

May 2014 – Jul 2014 · 2 mos

• Implementation of Motion Planning Algorithms (Eg: PRM, RRT etc) on robots and simulate them on Gazebo 3D Simulator running on ROS (Robot Operating System).