RadGuard AI
Designed and developed an AI-powered radiology second-read platform leveraging Google's CXR Foundation model to assist radiologists in detecting findings on chest X-rays. Architected the end-to-end ML pipeline including DICOM data ingestion, preprocessing, model inference, and structured output generation. Defined annotation schemas and ground truth labeling standards for training and evaluation datasets, with particular focus on edge case handling and inter-annotator consistency. Built the system with HIPAA-aligned data handling practices and designed the human-in-the-loop review workflow to ensure clinical-grade reliability. Positioned the platform for B2B deployment targeting radiology groups and teleradiology services seeking scalable second-opinion infrastructure.