Hameed Hasan

Developed a deep convolutional neural network for the prediction of threats (i.e., persons carrying guns), in real-time. This project was the analytics part of a platform developed by the company I had collaborated with. The overall platform included using a multi-static radar system that generated 3D volumetric images and a Kinect camera to find the distance to the camera. The radar phased data were then preprocessed and combined with the Kinect image to generate a 3D volume. We used state-of-the-art deep image segmentation and detection to uncover threats. This project involved the use of popular deep architectures such as R-CNN, Fast/er R-CNN, and Mask RCNN.

Worked on a pipeline based on Siamese networks to expedite the read alignment score, which can improve and accelerate the assembly process. The readings can potentially belong to different variants and genomes, and we were interested in making sure contaminated samples will not lead to inferior contig quality.

The objective of this project was to assess the compliance of submitted documents from healthcare providers with the required criteria and clinical guidelines. These guidelines serve as curated medical documents developed by doctors and clinicians to ensure that treatments are delivered according to the standards of care. The primary focus was to verify whether the providers had completed all the medically necessary tasks for any performed medical procedure, emphasizing the importance of adherence to these tasks.

The objective of this project was to design, develop, and train advanced transformer-based models capable of accurately answering medical questions. The approach employed retrieval-based techniques followed by question answering to extract relevant answers for a given query. By leveraging these methods, the project aimed to enhance the accuracy and efficiency of obtaining precise medical information in response to specific inquiries.