1. Design and implement secure cloud pipelines that ingest very large scan datasets (multi-terabyte), reliably and resumably
2. Build orchestration for GPU-accelerated reconstruction and analysis with strong retry semantics, idempotency, and cost controls
3. Define end-to-end data lifecycle for medical imaging: raw vs intermediate vs derived artifacts, retention policies, and reproducibility
4. Implement security + compliance primitives appropriate for HIPAA/PHI: encryption in transit/at rest, key management, least privilege, audit logs, and access reviews
5. Build operational tooling: monitoring, alerting, runbooks, and incident-driven improvements for a growing device fleet
Building reliable data pipelines at scale (queues/orchestration, resumable uploads, GPU batch execution) with strong observability
Security + privacy by default: encryption, least-privilege access, auditing, and practical HIPAA/PHI guardrails
Owning the “boring” backend details that keep a lean team moving: schemas/migrations, cost controls, retries, and runbooks
Understanding compute tradeoffs across hardware options, and specifying appropriate cloud resources