Presentation: Security and Architecture: To Betray One Is To Destroy Both

Security and architecture aren't separate concerns—they're interdependent systems that fail together. In my experience leading data platform migrations, I've seen teams treat security as a post-implementation compliance checkbox rather than an architectural constraint. This perspective is dangerous. When you decouple authentication and encryption from data pipeline design, you create fragile systems that collapse under real-world stress. The 'betrayal' framing is apt: ignore security in your dbt workflows, Kafka clusters, or feature stores, and your architecture becomes theoretically sound but operationally worthless. I've watched teams rebuild entire data warehouses because access controls weren't baked into partition strategies. The concrete takeaway is this—involve security architects in schema design, not just infrastructure reviews. Your data modeling decisions should encode least-privilege principles from day one. As organizations scale AI/ML workloads on shared platforms, this becomes non-negotiable. Teams treating security as a bolt-on rather than foundational will face either costly rewrites or governance failures when they hit production scale.