Mohit Kumar

I've built a real-time neural phase-retrieval workflow for clinical dark-field radiography, using a Physics-Informed Neural Network to address hardware micro-vibrations.

My Digital Talbot-Lau Alignment project used a modified U-Net in PyTorch, skip connections, and SSIM validation to retain high-frequency spatial data. I also designed an edge-inference pipeline that acts as a DICOM C-STORE SCP, processes unaligned X-ray data in under 50 milliseconds, and routes diagnostic maps to central PACS archives.

As a Clinical Radiography Intern in a Diagnostic Imaging Department, I executed daily imaging protocols, supported high-volume patient throughput, and worked with hospital IT systems and legacy PACS networks to keep radiological data accurately routed and stored.

Alongside clinical imaging, I deploy and troubleshoot local AI models on consumer hardware, create digital content, and produce technical documentation. I'm pursuing PACS Administrator and Medical Imaging Technologist opportunities where I can connect clinical workflows with healthcare IT infrastructure.