ONTOLOGY · v1.2

FDE / IRE Enterprise Skills & Lifecycle Ontology

A unified knowledge model for Forward Deployed Engineers, Inward Redeployed Engineers, AI-native skill governance, and tacit knowledge provenance — with a real-world showcase from Uber's Agentic Pods.

🚀 Forward Deployed Engineers 🔄 Inward Redeployed Engineers 🧩 9 AI Literacy Skills 📋 Skill Lifecycle Governance 🔗 PROV-O Provenance 🚕 Uber Agentic Pods Showcase

KG curated by rdf-infographic-skill and Claude Sonnet 5 on behalf of Kingsley Idehen

The Two Engineering Roles

The FDE/IRE distinction establishes a complementary boundary-spanning model — one role faces the world, the other faces the organisation, both governed by the same EngineeringRole class.

FDE · Outward-Facing
Forward Deployed Engineer
Embedded in customer environments. Deploys and adapts AI systems to real-world business contexts. Acts as the execution and integration interface between vendor systems and customer operations.
REQUIRES:
LLM Mastery Agent Engineering API Interoperability HTTP Mastery Data Space Tooling
IRE · Inward-Facing
Inward Redeployed Engineer
Extracts tacit knowledge from enterprise teams and transforms it into structured, reusable, AI-executable skill artifacts. Acts as the internal knowledge compression layer.
REQUIRES:
LLM Mastery Skill Engineering Linked Data Data Access Protocols Query Languages Agent Engineering

9 AI Literacy Skills

The AILiteracySkill class defines the capabilities that separate AI-native engineers from AI-adjacent users. All 9 are subclasses of Skill.

🤖
LLM Mastery
Mastery of frontier LLMs across vendors — model selection, prompt engineering, context management, and output evaluation. The foundation of all AI-native work.
Deployed
🕸️
AI Agent Engineering
Design of modular agents composed of skills, tools, and memory systems. Enables transition from prompt-based usage to autonomous, composable AI systems.
Deployed
🧩
Skill Engineering
Design of reusable skills decoupled from tools and data spaces. Transforms tribal knowledge into composable enterprise execution units.
Validated
🗄️
Data Space Tooling
Building CRUD tools across heterogeneous data spaces. Without this, AI cannot act on real enterprise data — it's just reasoning in a vacuum.
Deployed
🔌
API Interoperability
Design and consumption of OpenAPI and Model Context Protocol (MCP) interfaces — the connective tissue enabling cross-system AI orchestration.
Deployed
🌐
HTTP Protocol Mastery
Deep understanding of HTTP semantics, REST, content negotiation, and authentication. HTTP is the universal transport layer for AI-enabled distributed systems.
Deployed
🔗
Linked Data & Semantic Web
Modeling structured knowledge using RDF, ontologies, and SPARQL. Enables machine-readable enterprise knowledge and cross-domain reasoning.
Validated
🔑
Data Access Protocols
Database connectivity via ODBC, JDBC, and equivalents. Ensures AI accesses complete enterprise datasets — not just fragments exposed through higher-level APIs.
Deployed
📊
Declarative Query Languages
Mastery of SQL, SPARQL, GraphQL, GQL, and openCypher. The universal language of truth retrieval from heterogeneous enterprise data systems.
Validated

Skill Lifecycle

Every Skill artifact is tracked through four SkillLifecycleState governance stages. Skills are never deleted — they are deprecated and retained for audit traceability.

💡
Proposed
Under formation. Not yet validated for production use.
✅
Validated
Reviewed and approved for correctness and applicability.
🚀
Deployed
Actively used in production or operational environments.
📦
Deprecated
Retired but retained for traceability. Never deleted.

Provenance: From Tacit Knowledge to Skill Artifact

The PROV-O model captures how John (an IRE at Enterprise Business Org) transformed a Tacit Knowledge Fragment into a versioned skill artifact with full audit trail.

prov:Entity
Tacit Knowledge Fragment
→
SkillGenerationActivity
Codification Activity
by John (IRE)
→
Skill · ProposedSkill
KnowledgeCodification Artifact

How to Codify Tacit Knowledge as a Skill Artifact

The 5-step IRE workflow for transforming organisational expertise into AI-executable, lifecycle-governed skill artifacts.

1

Identify Tacit Knowledge Source

Locate and interview subject-matter experts, review undocumented workflows, and inventory processes that exist only in team members' heads. Use discovery interviews and process shadowing to surface hidden expertise.

2

Extract and Structure the Knowledge

Convert raw expertise into structured form using RDF, schema.org, and skill definition templates. Capture trigger conditions, execution steps, expected outputs, edge cases, and failure modes. Mint a document-local IRI and encode as Turtle or JSON-LD.

3

Validate with Domain Experts

Review the structured skill artifact with knowledge holders and peer engineers. Record the validation as a prov:Activity. Transition the artifact to ValidatedSkill state on approval.

4

Deploy as a Reusable Skill

Register in the enterprise skill registry. Configure for execution by AI agents and human engineers. Transition to DeployedSkill state. Publish the RDF artifact with a dereferenceable IRI.

5

Maintain Lifecycle Governance

Monitor effectiveness continuously. Update skills using prov:wasRevisionOf for version linking. Deprecate obsolete skills with DeprecatedSkill state. Never delete — deprecate and retain for audit.

Real-World Showcase: Uber Agentic Pods

A public example validating the IRE codification model at enterprise scale — from Praveen Neppalli Naga of Uber, via a LinkedIn post.

Uber Agentic Pods — agentic AI adoption showcase image from Praveen Neppalli Naga's LinkedIn post
IRE at Enterprise Scale
Agentic Pod
A cross-functional team pairing an AI engineer with a domain expert from finance, legal, marketing, or support. Each pod runs a compressed two-week cycle — shadow, prioritize, prototype, validate, ship — mirroring the IRE_Codification_HowTo steps at organizational scale. Uber reports 99% engineer AI-tool adoption, 70%+ of pull requests involving agents, and 2,500+ agent skills built across the development lifecycle.
💰
Capital Allocation
Cycle time compressed from 15 hours to 30 minutes after Agentic Pod workflow redesign.
📄
Financial Reports
Turnaround compressed from 2 days to 10 minutes.
🎧
Support Automation
9,000 manual support workflows converted to self-service automation.

Core insight: the workflow, not the individual task, is the unit of automation — real value comes from redesigning entire processes and eliminating handoffs and unnecessary approvals once engineers deeply understand how work actually flows across systems.

Frequently Asked Questions

Core concepts from the ontology, answered in plain language.

A Forward Deployed Engineer is an outward-facing engineering role embedded in customer environments. FDEs deploy and adapt AI systems to real-world business contexts, acting as the execution interface between an AI vendor's systems and the customer's operations. They require LLM mastery, agent engineering, API interoperability, HTTP protocols, and data space tooling skills.
An Inward Redeployed Engineer is an inward-facing role focused on extracting tacit knowledge from enterprise teams and transforming it into structured, reusable, AI-executable skill artifacts. IREs act as the internal knowledge compression layer. They require skill engineering, linked data, data access protocols, query languages, and agent engineering skills.
The skill lifecycle has four governance states: Proposed (under formation, not yet validated), Validated (reviewed and approved), Deployed (actively used in production), and Deprecated (retired but retained for traceability). Every skill artifact progresses through these states to ensure governance, auditability, and organisational trust.
Tacit knowledge is unstructured operational expertise distributed across enterprise teams — undocumented processes, expert intuition, and tribal know-how. IREs codify it through a SkillGenerationActivity: identify tacit knowledge sources, extract the expertise, structure it as an RDF-backed skill artifact, validate with domain experts, and deploy for reuse by AI agents and engineers.
AI Literacy comprises 9 core capabilities: LLM mastery, agent engineering, skill engineering, data space tooling, API interoperability, HTTP protocol mastery, linked data and semantic web, data access protocols, and declarative query languages. Together they determine whether an enterprise can move from passively consuming AI to designing AI-native operational systems.
FDEs face outward — they deploy AI in customer environments and generate new operational knowledge through that work. IREs face inward — they capture that knowledge and codify it into reusable skills. This creates a virtuous cycle: FDE deployments surface new tacit knowledge, IREs encode it as validated skills, and those skills make subsequent FDE deployments more effective and consistent.
An AI-executable Skill must encode: (1) trigger conditions — when it should be invoked; (2) execution steps — a precise, ordered procedure; (3) expected outputs — what success produces; (4) edge cases — known failure modes and resolutions; (5) provenance — who created it, from what knowledge, and when. Structure, determinism, and composability are the hallmarks.
Yes — Uber's Agentic Pods, described publicly by Praveen Neppalli Naga, operationalize the IRE knowledge-codification pattern at enterprise scale. Cross-functional teams pair AI engineers with domain experts in finance, legal, marketing, and support, running compressed shadow-prioritize-prototype-validate-ship cycles that mirror the IRE_Codification_HowTo steps. Results include capital allocation dropping from 15 hours to 30 minutes, financial reports from 2 days to 10 minutes, and 9,000 manual support workflows becoming self-service — validating the core IRE insight that the workflow, not the task, is the true unit of automation. See the Showcase section above.

Glossary

Key ontology terms with their definitions and linked knowledge graph IRIs.

Forward Deployed Engineer (FDE)
Outward-facing engineering role embedded in customer environments to deploy and adapt AI systems. subClassOf EngineeringRole.
Inward Redeployed Engineer (IRE)
Inward-facing role that extracts tacit knowledge and transforms it into structured, AI-executable skills. subClassOf EngineeringRole.
Skill
A reusable, structured unit of operational capability executable by humans or AI agents. The atomic unit of AI-native enterprise capability.
AI Literacy Skill
A subclass of Skill encompassing the 9 core capabilities required to build, operate, and govern AI systems at enterprise scale.
Skill Lifecycle State
A governance classification tracking a skill artifact through Proposed → Validated → Deployed → Deprecated stages.
Skill Generation Activity
A prov:Activity that transforms tacit or explicit knowledge into structured skill artifacts. Records who did the work and what knowledge was used.
Tacit Knowledge
Unstructured operational knowledge distributed across enterprise teams — undocumented processes, expert intuition, and tribal know-how awaiting codification.
requiresSkill
A property associating an EngineeringRole with the Skill instances that practitioners of that role must possess. Domain: EngineeringRole, Range: Skill.
Agentic Pod
A cross-functional team pairing an AI/agent engineer with a domain expert to identify, prototype, validate, and ship workflow automation within a compressed cycle. subClassOf SkillGenerationActivity.

Knowledge Graph Explorer

Interactive D3.js graph derived from the companion RDF. Click nodes or edge labels to explore entities via URIBurner. Drag to pin; double-click to unpin.

Loading…

⚙ KG Settings

Physics Simulation

Predicate Display

Edge Predicates

Node Filtering

Resolver

Arrow Style

Classes Properties Instances