Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Why IT project estimates are almost always wrong - and what to do about it
The gap between estimated and actual delivery time in IT projects is a known and persistent problem. A look at the structural reasons it keeps happening and the practices that reduce it.
Data breach: what to do in the first 72 hours
A practical breakdown of how companies respond to personal data incidents - and why most of them get it wrong.
API-first for internal systems: why it matters before you have many of them
Building internal tools and systems with an API-first approach is not extra work. It is the discipline that prevents the integration mess most companies spend years untangling.
Cloud vendor lock-in: how to make a conscious decision
When vendor lock-in in the cloud is a reasonable trade-off, and when it is a risk worth assessing upfront.
RPA: what it actually solves and where it hits a wall
Robotic process automation is useful in a specific set of circumstances and fragile outside them. A plain account of what to expect before committing to an RPA project.
Why feature engineering still matters in the deep learning era
Deep learning automates feature extraction - but it does not remove the need to think carefully about what data you feed into the model.
ML models decay silently - and most companies do not notice
A model that was accurate at launch will gradually stop being accurate as the world changes. Why monitoring for model decay is not optional, and how to set it up before it becomes an incident.
BERT and the new baseline for applied NLP
What the BERT model changes in the practical use of text processing, and why it matters for companies working with unstructured data.
On-premise vs cloud: most companies end up with both
The debate between keeping servers in-house and moving everything to the cloud rarely ends with a clean answer. A look at what a realistic hybrid posture actually involves.
Who owns data quality in a company that is not a data company
Data quality problems are common. Accountability for them is rare. A look at how to assign ownership without creating a bureaucratic layer that nobody uses.
The real cost of adopting Kubernetes
What companies fail to account for when deciding to move to Kubernetes: not just technical complexity, but organisational and staffing challenges too.
Streaming data: when you need it and when batch is enough
How to decide whether your company needs streaming data processing, or whether that is unnecessary complexity for tasks that batch loading handles perfectly well.