Notes on data, AI, IT
and security
No marketing fog. The way I think about real problems with founders and managers.
Data as a product: why you cannot put one team in charge of all the data
When analytics stops working, the problem is usually not the tools. How to distribute data responsibility across teams.
When you should not break up the monolith
Microservices sound modern, but decomposing a monolith without sufficient reasons creates more problems than it solves. How to think about this decision.
Autonomous mobile robots in the warehouse: running the economics
AMRs are no longer a future concept - they are a working tool. I look at when they pay off and when buying a robot turns out to be an expensive mistake.
The Capital One breach: the cloud is not to blame, configuration is
In July 2019 Capital One lost data on over 100 million customers. I look at what happened and why the main lesson is not about the cloud - it is about access management.
Why ML teams keep rewriting the same thing over and over
Feature stores and feature management in machine learning: where the duplication comes from and how to get rid of it.
GPT-2 and language models: what the signal means for business right now
After GPT-2, the conversation about text generation shifted. I look at what actually changes for companies today and what is still in the lab.
AutoML: what it is and what a manager should not expect from it
How AutoML tools lower the barrier to machine learning - and where they still require expertise and management decisions.
Cloud security after the first wave of containers: more than just the network
Why the perimeter security model does not work in a container environment - and what actually needs protecting instead.
Stream processing is an operations question, not an architecture question
Why the decision to move to streaming should start with understanding operational load, not with choosing a technology.
Cloud egress cost: the hidden budget line that is easy to miss
Why the cost of transferring data out of the cloud often comes as an unpleasant surprise - and how to keep it under control.
Autonomous mobile robots in the warehouse: when it makes sense and when it does not
How to assess warehouse readiness for AMR deployment - without the marketing simplifications that robots will fix everything.
The real cost of an NLP pipeline before you are sold by the demo
What actually requires ongoing support in a production NLP system - from data labelling to quality control in live operation.