FDE role and project qualification

Qualification · Stakeholders · Fit/gap · Environment

  • Translate AI ambition into a broken job
  • Four gates for project fit
  • The delivery–product boundary
  • Make the first commitment testable

Read module ↗

From fieldwork to lasting business capability.FORWARD-DEPLOYED ENGINEERING

CURRICULUM / FDE PRACTICE

Twelve modules follow the engagement. The expanded Chinese notes include explanations, engineering examples, cases, assignments and review criteria; weeks five and nine are technical deep dives.

Qualification, discovery, metrics and scope

Adapters, ontology, tools and data boundaries

Evaluation, controlled actions and operations

Attribution, retrospective, assets and defense

Twelve weeks is a learning structure, not a guaranteed deployment timeline. Allow time for implementation, model integration and review; reading progress is not assessment.

01 PHASE 1

Output: Fit memo + fit/gap register + run evidenceRead module ↗

02 PHASE 1

Interviews · Evidence ledger · Stakeholders · Ontology draft

  • Interview an incident, not an aspiration
  • Separate facts, interpretations and incentives
  • Map influence and actual involvement
  • Model nouns, relationships and actions

Output: Interview ledger + stakeholder map + ontology v0Read module ↗

03 PHASE 1

Metric contracts · Executable definitions · SOW · Change

  • Connect the north star to an actionable job
  • Eight fields in a metric contract
  • Prioritize by evidence and constraints
  • Scope of work is a shared understanding

Output: Metric contract + scope draft + executable metric testsRead module ↗

04 PHASE 2

Profiling · Quarantine · Data contracts · Thin-slice demo

  • Profile defects that change decisions
  • Treat the adapter as a contract boundary
  • Prove one data chain before a full platform
  • Make the demo reveal useful disagreement

Output: Data diagnosis + adapter + first replayable sliceRead module ↗

05 PHASE 2DEEP DIVE

Identity · Semantic layer · Retrieval · Advisory tools

  • Entity resolution: examine false merges
  • Ontology makes data actionable
  • RAG and metrics answer different questions
  • MCP connects tools; it does not define the task

Output: Ontology and semantic config + tools + advisory MVP design and tracesRead module ↗

06 PHASE 2

Data flows · Field access · Utility tests · Deployment

  • Trace the complete data journey
  • Masking, encryption and anonymity differ
  • Enforce permissions outside the model
  • Cloud, private and hybrid are tradeoffs

Output: Security design + permission tests + deployment decisionRead module ↗

07 PHASE 3

Datasets · Graders · Regression · Release gates

  • Evaluate the system completing the task
  • Three test sets protect different boundaries
  • Use complementary graders
  • Turn objections into the next test

Output: Evaluation sets + regression report + change proposalRead module ↗

08 PHASE 3

Coupons and tags · Confirmation · Idempotency · Feedback

  • An action is a precise contract
  • Separate autonomy from risk
  • Retries, rollback and compensation differ
  • Adoption comes from fitting the job

Output: Action contract + failure tests + adoption ledgerRead module ↗

09 PHASE 3DEEP DIVE

Guardrails · Observability · Rollout · Cost · Handover

  • Six dimensions of readiness
  • Guard inputs, tools and outputs
  • Use traces to locate failures
  • Design rollout, economics and handover together

Output: Launch pack: checks, runbook, cost model and canary planRead module ↗

10 PHASE 4

Acceptance · Observation windows · Controls · Outcomes

  • Three evidence layers are not interchangeable
  • Design attribution before calculating it
  • Released hours are not automatically cash savings
  • Outcome-linked acceptance needs shared boundaries

Output: Attribution analysis + acceptance pack + outcome boundariesRead module ↗

11 PHASE 4

Failure modes · Asset layers · Product feedback · Transfer

  • A retrospective that preserves evidence
  • Audit six delivery failure patterns
  • Platform, business agents, FDE and enablement
  • Test reuse with a second scenario

Output: Retrospective + asset layers + transfer playbookRead module ↗

12 PHASE 4

Reproduction · Stakeholder review · Decisions · Portfolio

  • Hand the project to another person
  • Defend why, not only what
  • Assess competence, not certificate theater
  • Make growth and portfolio claims verifiable

Output: Capstone pack + defense materials + growth recordRead module ↗

No matching modules.