• HeartFlow-Imperial College London collaboration on deep learning-based CCTA reconstruction and multi-phase image registration.

• Conception, development, and training of deep learning (DL) image registration models on large 3D multiphase coronary computed tomography angiography (CCTA) datasets for construction of cardiac motion models suitable for simulation-based data augmentation of CT acquisition and reconstruction.

• Conception, development, and training of DL coronary artery motion quantification from single 3D CCTA images to detect and quantify cardiac motion from CT reconstruction motion artifact signature.

• Implemented CUDA-accelerated PyTorch operator for differentiable B-spline deformation field sampling.

• Implemented tools to parse and curate multiphase CCTA DICOM files.

• Developed interactive, point-to-surface 3D image annotation tool built on Slicer 3D for manual segmentation of cardiac anatomy.

• Led transition of Image Analysis team deep learning framework to TensorFlow.

• Thought leader for cross-functional proposal to revamp Production Image Quality Assessment process to produce ‘learnable data’ to realize virtuous cycle of automation, learnable human correction, automation improvement.

• Scrum master for Image Analysis sub-team.

• Training and supervision of new hire algorithms engineer.

• Reformulated TensorFlow CMake build system to support dynamic linkage against internal super builds of TensorFlow dependencies.

• Implemented TensorFlow support for user-encrypted model persistence leveraging TensorFlow’s memory mapped filesystem.

• C++ development support for next generation, deep learning-based centerline extraction and lumen segmentation.

• Development and application of active shape model framework for cardiac chamber segmentation.

• Artifact detection in cardiac computed tomography angiography images (motion, blooming, misalignment) to support Production Image Quality Assessment process.

• Design and implementation of tools to parcellate Left Ventricular Myocardium segmentation according to AHA 17-segment model.

• Product management, design, implementation, and testing of C#/WPF-based frontend intermediary between Production Workstation for review and correction of image analysis automation results and Production Backend.

• Workflow design for development of next generation Production Workstation.

• Participation in pre-commercial case processing for DEFACTO clinical study.

• Evaluation of coronary artery lesion segmentation and centerline extraction algorithms.

• Generation and curation of product design documentation for 510K FDA filing.