FDE glossary

The terms I use in engagements and cohorts, defined in one or two sentences each. Where a term has a common meaning elsewhere, this is the meaning I use.

Terms

Forward-deployed engineering (FDE)

A way of delivering technology in which engineers work inside the customer's environment, close to the problem, and stay until the result runs. The engineer owns an outcome, not a backlog.

Forward-deployed engineer

An engineer who does forward-deployed engineering. Needs two things: technical depth and customer judgement. Also abbreviated FDE.

Deployed team

A small group of forward-deployed engineers placed with one client, with a named lead who is accountable for the result.

Discovery brief

A one- or two-page record of the task, its owner, the evidence that matters, the data and systems that can be reached, and what is out of scope. The output of the first gate.

Architecture decision record (ADR)

A short document that records one technical decision, the options considered and the reasoning. A set of them is the output of the second gate.

Evaluation harness

The code and test cases that measure whether an AI system does the task to the agreed standard, run repeatedly as the system changes.

Responsible AI controls

The evaluation, human review steps, data boundaries and audit records that are written down before a build starts, so the system can be trusted and checked after launch.

Handover

The point at which a named owner inside the client takes over running the system, with a runbook, the evaluation results and the recorded decisions. The output of the fourth gate.

The four gates

My assessment structure for forward-deployed work: G1 context and problem framing, G2 architecture, G3 integration, G4 defend and hand over. Each has a written output and a rubric. See the framework.

AI Delivery Readiness Review

My two-week advisory engagement on one workflow or pilot, ending in a written recommendation to proceed, reframe or pause. See Engagements.

Architecture Sprint

My two-to-four-week engagement that turns an agreed use case into a solution architecture, recorded decisions, an evaluation plan and a delivery backlog.

Retrieval-augmented generation (RAG)

A pattern in which a language model is given relevant documents or records at the time of the question, so answers rest on the organisation's own data rather than on the model's memory.

Agent

A system in which a language model decides which tools or steps to use to complete a task, within limits set by the designer. In enterprise work the limits and the review steps matter more than the model.

Capability centre

A unit inside a services firm or large enterprise, often in India, that builds and staffs a technical capability for the wider organisation. Many of my cohorts are for capability centres.