Forward Deployed Engineering
Forward Deployed Engineering: from model to enterprise
A working model is not a deployed decision system. The missing work is provenance, canonical mapping, coupling, tuning, ownership and a method another team can operate.
9 minute read
Applied AI in production
Applied AI in production: what the operating model requires
Production requires a contract: what the organisation declares, what the system derives, who owns the result and how the method can be challenged and rerun.
8 minute read
Benchmark audit
120,000 inferences: the economics of purpose built AI infrastructure
The latency result held. The economics claim did not. This is the audit trail, the measurement defect and the narrower result that survived.
8 minute read