Case studies
How the work went
Long write-ups of real engagements. Each one covers what was wrong, how I measured it, what I tried and rejected, and what couldn’t break while it changed. It also says where its numbers come from and what I left out.
~70% faster end to end
Biotech · Data platform
Five separate problems behind one API's timeouts
A scientific data API was failing with timeouts and 5xx errors. Behind them were five separate problems in the schema, query planning, transport and server concurrency, and from the outside they all looked the same.
Figures from measurements committed to the client's repository. Client and partner anonymised.
125 vendor actions, one dispatcher
Integrations · Open source
Covering 125 SaaS API actions with a contract built for databases
An integration contract designed for CRUD over a table had to cover accounting and commerce APIs. One reflective dispatcher handles about 125 vendor actions, and the connectors came with test suites in a catalog that mostly shipped without them.
Public MIT-licensed repository. Every figure is a file-level count or a dated commit.
Recognise one of these?
Most of my work starts with a system that’s slow and fragile, and nobody can say why. Thirty minutes is usually enough to tell whether I’m the right person for yours.
Book a 30-min intro call Free 30 minutes. No pitch, and no paperwork until we both want to continue.