About me

Hello and welcome!

I worked as a Senior Research Scientist at Augrade Inc., USA, on Generative AI and Computer vision applications at Augrade for a brief period. Prior to this, I worked as a Computer Vision and Image Analytics Research Scientist at Siemens AI, USA. In addition, I am a Senior Research Associate at the University of Southern California, Department of Civil Engineering, under Professor Sami F. Masri. Lastly, I am a reviewer for the Sage Structural Health Monitoring, Elsevier Automation in Construction, Computer Physics Communications and Springer Nonlinear Dynamics journals.

I obtained a Ph.D. in civil engineering, emphasizing applied computer vision and machine learning applications for vision-based methods in the civil engineering domain. In my previous Ph.D. research experience at the University of Southern California, I developed crack segmentation and change detection methods for civil and mechanical infrastructures using classical and deep learning methods. As a senior research scientist, I spearheaded a large-scale synthetic dataset generation pipeline for architectural, engineering, construction (AEC), structural, and MEP drawings to augment and train deep learning models. Furthermore, I have developed Generative AI solutions for the 3D CAD model generation and digital twin creation using the 2D drawings of AEC. In addition, I have developed on the graph neural networks for the hierarchical assembly graph definition, parametrization, and reconfiguration. Additionally, I have developed methodologies for synchronizing industrial digital twin models using the 3D point clouds, images, and other modalities of datasets (images, and RGB-D) during my tenure at Siemens, USA. Recently, I drove two projects at the University of Southern California on synthetic structural cracks generation algorithms that can assist generative paradigms like Generative Adversarial Networks (GANs) or Denoising Diffusion Probabilistic Models (DDPMs) and Latent Diffusion Models (LDMs) to facilitate unpaired image-to-image translation and semantic mask synthesis, generating diverse and unique binary crack masks. Consequently, reducing the computational time (on average ~1.5 seconds to generate transverse, longitudinal, shear, branched, multi-branched, and fatigue or block cracks) and diversifying the synthetic cracks generated in contrast to physics-based computer graphics and physics-based finite element formulation of material models. Overall, my background lies in generative AI, computer vision, deep learning, machine learning and computational methods for civil engineering and large-scale industrial applications in general.

As a researcher, my focus is on both data-driven / machine learning methods in structural health monitoring problems, digital twins creation, synthetic datasets generation, computer vision, deep learning (Generative AI, Transformers, convolutional neural networks, recurrent neural networks, and neural networks in general), image processing and machine learning-based image recognition applications (such as semantic segmentation, object detection and classification of structural defects. Change detection and quantification of structural components). My expertise lies in concrete, pavement, and steel infrastructures and materials in general. I have domain expertise in multimodal data, e.g., color, depth images, and point clouds.