Platform Architecture

Responsibilities:

lead the design and implementation of an AWS ECS-based containerization platform for the company’s arterial analysis software, reducing production deployment time from hours to under five minutes
architect Docker images for the platform’s GPU workloads: PyTorch inference models with CUDA support, and a C++ rendering service built on VTK, EGL/OpenGL, and the CUDA runtime
author the reference Docker image for the platform’s Python/Django backend, then lead development of the C++ build and runtime images for processing and rendering CT imagery
design GitHub Actions workflows that build the C++ images and test them on GPU runners
architect a cross-region warm-standby disaster recovery footprint, defining the Terraform IaC patterns for network, compute, storage, and identity layers
design asynchronous task-processing infrastructure for backend services, introducing managed message brokering and dedicated worker tiers to support scalable, long-running workloads
develop the HTCondor cluster that schedules inference and rendering jobs onto auto-scaling GPU instance pools
guide the evolution of the company’s service-based architecture, partnering with engineering leaders to navigate organizational and technical trade-offs
build GPU-enabled base AMIs with EC2 Image Builder, and establish cross-account, multi-region distribution patterns for AMIs and container images, decoupling build pipelines from deployment topology
own multi-account cost management for over eighty environments serving six product development teams, balancing engineering velocity with budget accountability
develop DICOM interoperability and performance-testing tools, including a cross-platform C-ECHO service class user and a Python-based load-testing harness for DICOM service class providers