Blog
Notes from the gap between research and production.
Embedding engineers in research teams: what works
Practical lessons on pairing engineers with researchers so that both move faster and the work arrives in production intact.
Security reviews for research systems: start early, stay small
How to bring security into research work without stalling it, and why the first review should be about data and access, not code.
What a research-to-production handover should contain
A short, concrete list of what a team needs in hand before a validated result becomes a system someone else operates.
The compute stack your experiments quietly depend on
Scheduling, storage and cost visibility decide how fast a research team can move, long before the model does.
Reproducibility is an infrastructure problem
Most irreproducible results are not a failure of discipline. They are a failure of environment, data versioning and tooling.
Research code is not production code, and that is fine
Why the gap between a working prototype and a dependable system is a different kind of work, and how to plan for it from the start.