Taha Rehah

• ● Define and execute the technical vision and cloud architecture of DeepEcho, an FDA-regulated SaMD providing AI-powered fetal ultrasound analysis.

• ● Lead cross-functional teams of engineers, ML researchers, and clinicians, bridging AI research, regulatory compliance, and real-world clinical workflows.

• ● Architect and oversee a Google Cloud Platform (GCP) infrastructure leveraging Kubernetes (GKE), TorchServe/Triton, Dataflow, BigQuery, and Redis for scalable, real-time AI inference.

• ● Design and maintain a secure, HIPAA-compliant cloud system, implementing authentication, authorization, encryption, audit logging, and patch management in line with FDA cybersecurity guidance and NIST 800-53 controls.

• ● Direct the full ML lifecycle — data acquisition, labeling, model training (PyTorch/MMDetection), evaluation, MLOps deployment, and monitoring — with CI/CD pipelines using GitLab, Cloud Build, and Terraform IaC.

• ● Lead the development of DeepEcho's real-time segmentation SDK for fetal ultrasound, optimized for both live stream inference and DICOM file processing.

• ● Collaborate with the QA/RA team to ensure ISO 13485 and IEC 62304 compliance, including software risk management, V&V protocols, and 510(k) submission documentation.

• ● Spearhead initiatives for foundation model research, data governance, and privacy-preserving training on large multimodal ultrasound datasets.

• ● Act as the primary technical liaison with Google Cloud, clinical partners, and regulatory consultants to align innovation with compliance and market needs.

• ● Mentor engineers in MLOps, medical imaging standards (DICOM, HL7, FHIR), and cloud security, fostering technical excellence across the organization.

• ● Directing and innovating models deployment initiatives, implementing cuting-edge algorithms and architectures in medical imaging and diagnostics.

• ● Implemented continuous integration and continuous deployment (CI/CD) practices for ML models, enabling rapid iteration and deployment of new model versions.

• ● Spearheaded an advanced analysis of Volvo's warranty claims using Large Language Models (LLMs), extracting critical insights about prevalent vehicle issues and customer feedback paterns.

• ● Seamlessly integrated and deployed state-of-the-art LLMs Google FLAN-T5, automating the extraction and categorization of warranty concerns, significantly optimizing the claim analysis process.

• ● Fostered key collaborations with Volvo's Quality Control and Data Engineering teams, translating research into actionable solutions, and paving the way for enhanced product reliability and customer satisfaction.