Visual AI and data platform

Project Hafnia

A regulation-compliant platform that turned video data into governed material for visual AI model development. I served as the primary cloud architecture authority from concept through production deployment.

Role: Principal Cloud Architect
Period: January 2022 – October 2024
Context: Milestone Systems × NVIDIA

The challenge

Visual AI development needs large, well-described datasets. Video also carries significant privacy, security, and regulatory obligations. The platform had to support ingestion, preparation, annotation, governance, and model training without treating those obligations as a later concern.

The visible data pipeline sat on a cross-cutting cloud foundation: multi-account governance, access controls, encryption, monitoring, and cost visibility.

Architecture

A governed path from video to model training

Open full-size SVG →

The diagram explains the principal service relationships without exposing private client configuration. It represents a production-deployed professional engagement.

Architecture contribution

End-to-end ownership: target architecture and decisions from concept through production.
Governed data: controls were designed into ingestion, storage, annotation, and training flows.
Platform foundation: Organizations, Control Tower, IAM, KMS, and CloudWatch supported the workload.
Cross-organization design: architecture aligned the collaboration and its operating boundaries.

Discuss AI and cloud platform architecture.

Start a conversation →