Work
Systems already running, described honestly.
Organisations are described without being named, at their preference. The engineering is described exactly as it happened, including the parts that were harder than they should have been.
National healthcare payer
Four systems of record that had never been joined
A configuration database, a data warehouse, a resilience platform and a document estate reconciled into one evidence-backed view of application resilience risk — across roughly 180 enterprise applications.
Read the case studyThe result worth knowing
Adding unread vendor disaster-recovery test reports as a fourth evidence layer moved critical controls off attestation and cut one application’s measured risk by about a fifth. Nothing in the environment changed.
Data software company
Azure spend and capacity in one operating picture
Billing joined to utilisation telemetry across an on-premise cluster and a large Azure footprint, so that right-sizing and consolidation could be argued from evidence rather than instinct.
Read the case studyThe result worth knowing
The cost conversation moved from a monthly total to a named list of workloads with owners, utilisation and a saving attached.
Clinical documentation
A clinical scribe where the audio never leaves the building
Speech recognition running on hardware the practice controls, with structured note generation on top — a mobile-first product built around a hard data-residency constraint.
Read the case studyThe result worth knowing
The transferable result is the pipeline split: the sensitive step runs locally, the reasoning step runs wherever it is best served.
Your environment is probably messier than these. That is normal.
Every one of these started with sources that did not join, systems that were supposedly inaccessible, and a record nobody entirely believed.