Why Research Code Collapses in Production
Coming soon
Notebooks and scripts work until you need reliability, scale, or someone else to run them. What breaks, why it breaks, and what changes when you're accountable for uptime.
Short, opinionated pieces by Efe Ertugrul about infrastructure, research systems, and what actually happens when you ship.
Coming soon
Notebooks and scripts work until you need reliability, scale, or someone else to run them. What breaks, why it breaks, and what changes when you're accountable for uptime.
Coming soon
Everyone treats metadata as an afterthought until queries fail, pipelines break, or compliance audits happen. Why metadata design is an engineering problem, not a documentation problem.
Coming soon
Most performance problems aren't in your orchestration layer. They're in your queries, your data model, or your API design. When K8s actually matters and when it's a distraction.
Coming soon
Training a model is 10% of the work. The other 90% is data pipelines, versioning, monitoring, retraining workflows, and making sure it doesn't fail silently. Real numbers from production systems.
Coming soon
How converting XSD to LinkML enabled cross-platform validation for bioimaging metadata. The engineering decisions, the tradeoffs, and why this work matters more than it sounds like it does.