GitHub

AI FACTORY DEPLOYED ENGINEER

Design your software factory.

Use AI-FDE to turn evidence about your engineering workflow into a reviewed Factory Definition. An accountable human Factory Deployed Engineer owns the design; Mission Control qualifies the runtime.

Evidence → Reviewed design → Runtime qualificationInternal alpha · Explore a synthetic workspace.
Open engagement workspace See how AI-FDE worksAssess readiness Build a Factory Definition
AI-FDE / Evidence workspaceSource → definition
Your engineering system
RepositoriesTeamsCI/CDBacklogStandardsContext
AI-FDE + accountable FDE
01Ingest02Model03Review04Design05Approve06Prepare
Factory Definition
ObjectiveWorkflowCapabilitiesAuthorityVerificationEconomics
Mission Control boundary
ValidateQualifyVersionExecuteVerifyRelease

The operating team

AI-FDE assists.
The human FDE remains accountable.

AI-FDE is the product. The Factory Deployed Engineer is the human role that owns architecture, judgment, authority, adoption, and outcomes. Mission Control owns governed runtime qualification and execution.

01 / Human lead

Factory Deployed Engineer

Owns the factory design and organizational outcome.

  • Discovery
  • Architecture
  • Factory Definition
  • Workflow decomposition
  • Trust boundaries
  • Evaluation strategy
  • Rollout
  • Organizational adoption
02 / Evidence product

AI-FDE

Discovers and designs from reviewed enterprise evidence.

  • Evidence ingestion
  • Candidate claims
  • Current-state modeling
  • Target workflow design
  • Economics
  • Readiness
  • Handoff preparation
03 / Control plane

Mission Control

Governs the runtime and preserves the operating record.

  • Authority
  • Orchestration
  • Execution
  • Evidence
  • Verification
  • Recovery
  • Approval
  • Economics

AI-FDE in the FDLC ecosystem

One product boundary at every stage.

FDLC defines the model. AI-FDE discovers and designs the factory. The human FDE owns the judgment. Mission Control governs qualification and execution.

  1. 01FDLC

    Defines the operating model.

  2. 02AI-FDE

    Discovers and designs it.

  3. 03Human FDE

    Owns it.

  4. 04Mission Control

    Governs it.

  5. 05Factory

    Executes it.

  6. 06Outcome

    Proves it.

How AI-FDE works

Evidence in. Governed intent out.

AI-FDE assists the human FDE through six explicit stages. It does not silently promote inference to truth or approved intent to runtime.

  1. 01

    Discover

    Ingest enterprise source evidence and identify a bounded workflow worth modeling.

    Input
    Evidence + baseline
    Output
    Source map
    Evidence
    Attributed documents and observations
  2. 02

    Define

    Extract candidate claims and model the current workflow without treating inference as truth.

    Input
    Source map
    Output
    Current-state model
    Evidence
    Claim-to-source lineage
  3. 03

    Assemble

    Design target workflow, capabilities, execution architecture, economics, and authority boundaries.

    Input
    Reviewed current state
    Output
    Draft Factory Definition
    Evidence
    Decisions, assumptions, and unresolved claims
  4. 04

    Qualify

    Review the definition against readiness, verification, policy, rollout, and rollback requirements.

    Input
    Draft definition
    Output
    Review decision
    Evidence
    Attributable human approval or rejection
  5. 05

    Operate

    Prepare approved definition and qualification material for Mission Control validation.

    Input
    Approved definition
    Output
    Typed handoff request
    Evidence
    Version, digest, provenance, and approval
  6. 06

    Transfer & Scale

    Use independently verified runtime outcomes to inform controlled definition changes.

    Input
    Mission Control outcome evidence
    Output
    Change proposal
    Evidence
    Baseline comparison + governed review

The primary design artifact

Begin with the factory required to produce the outcome safely.

AI-FDE does not begin by choosing a model. It helps the human FDE define what the factory must produce, what may act, what evidence must return, where people decide, and how value will be measured.

Explore Factory Definition Inspect the architecture
Factory Definition12 required concerns
OutcomeA governed factory lineSelect a component to inspect its responsibility.
01Objective

The measurable outcome the factory should produce and the constraints it must respect.

02Workflow

The source workflow, target workflow, actors, systems, exceptions, and known unknowns.

03Source lineage

The evidence and reviewed claims supporting each material part of the definition.

04Capabilities

The agents, tools, skills, models, SDKs, harnesses, and implementations required.

05Context

Approved knowledge, retrieval boundaries, freshness rules, and attribution requirements.

06Execution

The runtime, sandbox, Work Graph, orchestration, and recovery design.

07Authority

Policy, delegated permissions, approvals, escalation, and accountable human decisions.

08Verification

Independent acceptance criteria, evidence requirements, and failure handling.

09Economics

Budgets, baseline assumptions, attributable cost, and value measurement.

10Reliability

Operational limits, observability, reconciliation, rollback, and rollout controls.

11Outcomes

The technical, operational, and user results that will determine whether the factory worked.

12Readiness

Maturity, confidence, unresolved claims, and the conditions required before qualification.

The human-led engagement model

Three phases. One evidence trail.

AI-FDE supports the evidence and design work. The human FDE owns the decisions. Scale remains conditional on evidence from real operation.

  1. 01

    FDLC Readiness

    Understand the system and choose the right first factory line.

    Inputs
    • Workflows
    • Repositories
    • Architecture
    • Baseline metrics
    • Governance
    • Constraints
    Human + AI-FDE work
    • Maturity assessment
    • Opportunity discovery
    • Workflow decomposition
    • Baseline measurement
    • Factory Definition
    OutputA qualified Lighthouse candidate.
  2. 02

    FDLC Lighthouse

    Build one real governed production line and prove whether it creates value.

    Factory path
    • Factory Definition
    • Mission Control
    • Factory Version
    • Real WorkOrders
    • Independent verification
    Measure
    • Cycle time
    • Quality and defects
    • Attributable cost
    • Human effort
    • Verification reliability
    OutputEvidence that the operating model works—or evidence that it does not.
  3. 03

    FDLC Scale

    Turn a proven line into organizational capability.

    Inputs
    • Repositories
    • Teams
    • Workflow classes
    • Shared capabilities
    • Policies
    Human + AI-FDE work
    • Evaluation
    • Economics
    • Operating controls
    • Reuse boundaries
    • Scale qualification
    OutputA governed path to expansion, conditional on Lighthouse evidence.

What AI-FDE produces

A governed definition—not an unreviewed recommendation.

The primary output is an inspectable, attributable Factory Definition. Runtime qualification and execution remain separate Mission Control responsibilities.

01

Factory Definition

A complete, reviewable specification for the line.

02

Current-state model

An evidence-backed description of how the selected workflow operates today.

03

Authority model

The people, permissions, decisions, and escalation boundaries the factory requires.

04

Qualification requirements

The tests and evidence Mission Control must require before execution.

05

Verification contract

The evidence required before state may advance.

06

Outcome baseline

The before-state against which value is measured.

07

Economics model

Attributable cost per attempt and accepted outcome.

08

Rollout and recovery plan

Roles, exceptions, rollback, and day-two operating intent.

09

Capability requirements

The agents, tools, profiles, and policy implementations to qualify.

10

Handoff material

Digest-bound approved intent for validation by Mission Control.

Good lighthouse candidates

Start where evidence is easiest to prove.

The first line should be commercially relevant and operationally bounded. It should teach the organization how to trust the factory—not ask the organization to trust everything at once.

Dependency modernization

Upgrade libraries and frameworks with compatibility proof.

API migration

Replace deprecated interfaces with supported patterns and contract tests.

Vulnerability remediation

Find, fix, test, and verify known vulnerability classes.

Quality remediation

Repair flaky tests, coverage gaps, and recurring static-analysis debt.

Documentation maintenance

Detect stale technical documentation and validate it against source.

Code review assistance

Produce evidence-backed findings while reviewers retain decisions.

Test modernization

Move brittle suites toward stable, meaningful behavioral proof.

Framework migration

Migrate repeated application patterns with explicit compatibility gates.

CI/CD remediation

Repair repeatable delivery failures with observable acceptance criteria.

Good first factory lines have

  • Bounded scope
  • Repetitive work
  • Clear acceptance criteria
  • Existing baseline
  • Independent verification
  • Measurable business value
  • Safe rollback

Before / after

Make the handoff explicit.

Before AI-FDE
  1. Scattered source material obscures the current workflow.
  2. Unstated assumptions become implementation decisions.
  3. Teams start work without agreed acceptance criteria.
Engineering starts from uncertain intent.
After the AI-FDE handoff
  1. AI-FDE links source evidence to reviewable claims.
  2. The human FDE approves a Factory Definition.
  3. Mission Control qualifies a Factory Version before execution.
Reviewed intent. Qualified execution.

Define proof before execution

Compare the future line with the baseline it replaces.

AI-FDE captures a measurement contract, not invented improvement claims. Real values appear only after Mission Control produces attributable execution evidence.

01

Cycle time

BaselineFactory outcome
Measured, not assumed
02

Human engineering minutes

BaselineFactory outcome
Measured, not assumed
03

Defect escape rate

BaselineFactory outcome
Measured, not assumed
04

Verification pass rate

BaselineFactory outcome
Measured, not assumed
05

Cost per accepted WorkOrder

BaselineFactory outcome
Measured, not assumed
06

Cost per released outcome

BaselineFactory outcome
Measured, not assumed
Inspect the measurement contract

Organizational transformation

From developers to builders.

FDLC expands the builder population beyond people writing every line of code manually. Builders express intent, constraints, acceptance criteria, and authority while governed digital workers execute bounded work. Managers increasingly lead humans and digital workers.

IntentConstraintsAcceptance criteriaAuthority

The FDE

One role across the whole system.

A Factory Deployed Engineer combines the disciplines required to make the line real and adoptable.

  • Software architecture
  • AI / agent engineering
  • Platform engineering
  • Developer experience
  • Evaluation
  • Security
  • Change leadership
  • Operating-model design
PeopleProcessArchitectureFactoryEvidenceOutcomes

How the human FDE works with you

Your domain. Reviewed evidence. Accountable decisions.

The human FDE leads the engagement with your experts while AI-FDE makes evidence, claims, decisions, and implementation intent inspectable.

01 / You bring
  • A real workflow
  • A baseline
  • Domain experts
  • Engineering constraints
02 / We build together
  • Factory Definition
  • Lighthouse
  • Evidence
  • Operating model
03 / You keep
  • The factory line
  • The evidence
  • The capabilities
  • The operating playbook
  • The ability to scale

The customer product

This is where your team runs the factory.

Mission Control is the operating dashboard delivered with the factory line. Your team sees work that needs attention, inspects bounded execution, reviews evidence, makes consequential decisions, and retains the audit trail.

Mission Control / Command CenterSoftware Factory Demo
Mission Control Command Center showing ranked work that needs operator attention

What these screens show: the deterministic Software Factory demo and its operator experience. They demonstrate product structure—not a live customer engagement, customer activity, or performance result.

Explore the full Mission Control product

Autonomy without abdicating control

See what ran, why it advanced, and how to recover.

Mission Control gives the team visibility into work, evidence, approvals, cost, and rollback. Agents may propose. Workers may execute. Verifiers may evaluate. Humans retain consequential authority. The control plane makes those boundaries enforceable.

Explore Mission Control
Mission ControlGoverned runtime
  • 01WorkOrders
  • 02Attempts
  • 03Factory Versions
  • 04Execution Profiles
  • 05Evidence
  • 06Verification
  • 07Human Authority
  • 08Recovery
  • 09Economics

Execution complete ≠ Verified ≠ Approved ≠ Released ≠ Outcome proven

Design partnership

Bring one real workflow and its current baseline.

Begin with the maturity assessment and Factory Definition. Direct intake remains closed until the engagement and privacy paths are ready; the preparation tools are available now.

Assess your factory readinessBuild a Factory Definition

Current AI-FDE status

Synthetic product path qualified. Live operation not claimed.

The public demo is deterministic and synthetic. Its source and browser gates are current, but this is not customer-operation, live-model, or Production-runtime evidence.

Product stageInternal alpha
Public pathSynthetic demo
Live qualificationNot activated
Live model calls0

Evidence boundary

Qualified source
Exact merged revision; full trust, contract, build, browser, and accessibility gates passed 2026-09-14.
Demo data
Deterministic synthetic fixtures only. Browser-local demo state.
Network boundary
Invalid fallback API origin; hosted demo must not make live service calls.

Factory Definition handoff

AI-FDE may
prepare qualification material · request validation · report validation status
AI-FDE may not
create a Factory Version · qualify a Factory Version · activate runtime execution
Mission Control owns
validation · Factory Version qualification · WorkOrders · Attempts · execution evidence · release authority

What should I do next?

AI-FDE helps design the Factory Definition that Mission Control can qualify.

Build a Factory Definition →Inspect Mission Control →Assess readiness →