Not a list of services: these are the questions I find myself
thinking about outside work. If one of them is on your mind too,
I’d be glad to compare notes.
01
Data platforms that outgrow their first design
Most data systems are designed for the volume they had on day one. The interesting part starts when writes, tenants and questions grow at different speeds and the original shape stops fitting.
When is it worth decoupling compute from storage, and when is it just more moving parts?
How far can a single PostgreSQL instance go with the right partitioning before something else is needed?
What does a deployment have to prove before anyone can trust that it actually happened?
A dashboard is only as good as the definitions behind it. I like the problems where data quality, semantics and testing meet, and where the cheapest fix is often something the system already produces.
How do you catch data that is wrong but looks perfectly plausible?
Can a data contract double as the documentation of what each field means?
When does a team really need a warehouse, and when are the files it already writes enough?
Agents are easy to demo and hard to trust. The questions I keep coming back to are about evaluation, permissions and memory: how to know an agent is doing well, and how to make sure it cannot do harm.
How do you check that an LLM judge agrees with the person whose standard matters?
Which capabilities should an agent never hold, rather than be told not to use?
How can an agent learn from its mistakes in a way a person can read and reverse?
I came to data from mathematics, and I still enjoy problems where a model, a proof or a bit of statistics changes the answer: optimisation, probability, measurement and anything where structure hides under noise.
Where does a simple statistical test replace an expensive system?
Which everyday business questions are optimisation problems in disguise?
How do you measure something when there is no ground truth to compare it against?
How this works
There is no offer or price list here. If you are working on something
that touches one of these questions, a side project, an open-source tool, a piece of
research or simply an idea you want to think through, write to me. I read everything and
reply when I have something useful to add.