Ivan Caban

Built and deployed transformer-based clinical NLP and RAG pipelines for automated medical record summarization and question answering, supporting HL7/FHIR interoperability and HIPAA-compliant access controls. Automated MLOps with Kubernetes, Docker, GitLab CI/CD and Airflow for continuous deployment, drift detection, and scheduled retraining.

Developed distributed model serving and training pipelines using TensorFlow Serving, gRPC, and Envoy, and deployed multi-region GCP clusters with autoscaling and tracing. Refactored preprocessing with Apache Beam to reduce pipeline latency by 45% and implemented feature caching and monitoring for real-time ML workflows.

Built scalable backend microservices and clinician-facing web applications to enable real-time inference for symptom classification and digital therapeutics, and engineered Kafka/Spark streaming pipelines for sensor and EHR data. Automated ML pipelines with Airflow and implemented CI/CD and monitoring to ensure secure, HIPAA-compliant deployments.

Developed backend services and real-time image processing pipelines for mobile ultrasound applications, leveraging OpenCV and microservices to improve image clarity and enable cloud-based synchronization. Implemented scalable, HIPAA-compliant deployments on AWS and on-premise systems to support clinical workflows.

Assisted in building machine learning pipelines for medical image analysis, preprocessing DICOM datasets and improving data augmentation workflows to enhance CNN training. Supported backend API development to integrate models with clinical systems and contributed to testing and deployment readiness for hospital environments.

Completed a Bachelor of Science in Computer Science at The University of Melbourne from August 2006 to April 2010.