Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Attack through a software vendor: when your perimeter starts elsewhere
Why a compromise of third-party software is a threat to your infrastructure, and how to think about managing this risk.
A data lake without governance is a swamp, not an asset
Why a data lake without access policy and governance turns into unmanageable storage that nobody can get trustworthy data from.
Data ownership matters more than a data platform
Why companies buy expensive data platforms and end up with the same problems - and what needs to be resolved before choosing a tool.
Model drift: why an ML system degrades without visible failures
Machine learning models in production lose accuracy over time - quietly, with no errors and no alerts. What drift is and how to monitor for it.
GDPR: nine months in and the first major fine
In January 2019 Google was fined 50 million euros under GDPR. What it means and why having a privacy policy is not the same as actual compliance.
From hype to inference cost: why AI must be measured as a production function
How to move from evaluating AI by its demo effect to evaluating it by the real economics of running a model in production.
Breaking up a monolith: why sequence matters more than speed
How to plan a migration away from a monolithic architecture without halting operations - on sequencing, risks, and rollback points.
Kubernetes operators: what the model is and why it matters now
What the operator pattern in Kubernetes is, why it has become the central way to manage complex applications in a cluster, and what that means for architecture.
API versioning is contract management, not a technical formality
Why every API needs a versioning policy, and what happens to integrations when there is none.
An ETL pipeline is a production line - monitor it accordingly
Why ETL failures are an operational incident, not a technical glitch, and how to build visibility into data flows.
Why AI projects die before they produce results: five recurring patterns
An analysis of the typical reasons AI initiatives stall or fail to deliver their promised impact - and what to do about it.
ML in production demands process before MLOps has a name
Why companies that are serious about machine learning inevitably reach the need to version not just code, but data, models, and experiments.