Hello friends, this is Santos. I'll be presenting on scaling medical imaging infrastructure for life critical healthcare applications. We describe how advanced flexible endoscopy uses this architecture. The agenda platform architecture, fundamentals. What is the core infrastructure? Real time video processing pipelines. What are the storage? Arctic architecture design principles. Then performance engineering and availability. What are the scaling strategies, optimization techniques, and how do you recover? From disaster during a procedure. What are the data management and integration tools we use? How are they connected to the cloud? What is the security and compliance we are following? Then we end with case studies and future directions. See when the medical platform fail, patients are directly impacted. The momentary system. Lag in bronchoscopy could mean missing a critical airway lesion that could be even cancerous. A storage failure might result in lost or lost diagnostic imagery that cannot be recreated. So more what are the technical challenges? So the stream bandwidth, the single procedure produces massive amount of uncompressed data sets. The latency, the end-to-end photon latency should be less than 200 milliseconds in best case scenarios, we have systems with 90 to one 20 millisecond latency. The uptime is 99.99%. We have multiple imaging modalities, RGB, near I fluorescence. Each have distinct data streams with unique processing requirements. Most of the modern endoscopes are at least HD requiring robust infrastructure. Core infrastructure, architecture, compute requirements. We need clusters with hundreds of CPU cores and multiple high-end GPUs for real time processing. Dedicated resources for video encoding transcoding Andys enhancement operations, separate compute pools for real time versus batch processing workloads. What is the network architecture? 10 gigabits per second. Backbone capacity, quality of service mechanism, prioritizing endoscopic traffic, multiple edge ingestion points with automatic fail, fail or capability. Specialized low latency parts for real time clinical workflows. The real time video processing pipeline, data integration, image enhancement. Apply advanced image processing algorithms for noise reduction, enhancement, color correction, et cetera, encoding compression H 26 5 to reduce bandwidth while maintaining quality distribution, multi-protocol streaming to viewing stations and adaptive bitrate capabilities. Tired storage, architecture. Hot tire, N-V-M-E-R-A active recent examinations with sub millisecond access latency. Warm tire is SSD storage. Recent procedures up to 30 days. Cold tire is historical procedures with adaptive or intelligent compression algorithms. For cost effective long-term retention, horizontal scaling and performance optimization. Stateless microservice design, enabling linear capacity expansion, Kubernetes based or orchestration with automatic port scaling based. Intelligent load balancing, considering data, locality, and network topology. Session affinity, ensuring consistent processing for procedure streams. Performance optimization. GPU Acceleration for compute intensive video processing zero copy buffer management, minimizing memory transfers. ADD caching for frequently accessed content time series databases, optimized for procedure metrics, high availability and disaster recovery. What is the redundancy architecture? Eliminate single points of failure through redundant power network. Automated failover, continuous health monitoring with sub sub millisecond detection of component failures. Geographic resilience, comprehensive chaos engineering, and regular failover testing are essential. Partial failure test, partial failure modes, network partitions and cascading failures. Scenarios under control, conditions to validate performance, resilience, data management, analytics, infrastructure, realtime ingestion, processing framework analytics platform BV. Propose using a Kafka buffering base, high batch, high bandwidth video feeds, stream processing for initial validation and enrichment. For the processing framework, it is lamb architecture, combining stream and batch processing, GPU acceleration for computationally intensive tasks analytics platform. It's self-service analytics for quality improvement. Feature stores for consistent model training, security and compliance framework, HIPAA compliance, a S 2 56 encryption. For data addressed hardware security modules, role based attribute control systems, tamper proof audit, logging with cryptographic signatures, data privacy protection, end to end encryption from capture through Archer, through archive hierarchical key management with perfect forward secrecy. Data classification. Classification, driving automated protection policies, access control systems. Context aware authorization, multifactor authentication with single sign-on integration, behavioral intering, suspicious access patterns, healthcare system integration, single sign on. Federated architecture as reference pointers, PACS connectivity, DICOM interfaces converting video to multi frame objects, intelligent routing, balancing immediate availability with source storage. DICOM web WA interface for restful access. XT XTS frameworks enabling cross enterprise sharing. Cloud and hybrid deployment strategies. Cloud native architecture kubernetes authorization, enabling sophisticated deployment patterns, hybrid implementation, edge computing within the hospital networks for latency sensor operations, multi-cloud resilience, abstraction layers, enabling portable deployment across providers. Case study, academic medical center implementation. The, we had a deployment profile of 1200 bed academic medical center performing 50,000 endoscopic procedures requirements, 99.99% uptime, subsecond latency, a Kubernetes ization enabling dynamic scaling during peak procedure times. The key results were reduced latency to under 200 milliseconds while providing virtually unlimited storage EMR integration with federated architecture, preserving video quality, successful handling of 200 plus concurrent procedures during peak time, about 40% reduction in op operational cost through performance optimization. Future directions, AI and ML integration, edge computing advanced imaging for high definition, ultra high definition eight K 3D reconstruction from endoscopic video enabled enhanced visualization navigation assistance. Thank you.