AI-Native Data Product Design Standards

A comprehensive architectural framework for building AI-Native Data Products on Teradata — modular, self-describing, and optimised for autonomous agent discovery and operation.

---

Overview

This repository contains the master set of design standards for AI-Native Data Products. The standards define a six-module architecture where each module has a distinct responsibility, its own data model, and integrates with the others through consistent patterns.

Design documents are the single source of truth. The unified Claude skill (generated from these documents via the conversion prompt) is a compressed, agent-optimised rendering of the standards — never edited directly.

---

Repository Structure

ai-native-data-products/
├── design-standards/        ← master source of truth
│   ├── AI_Native_Data_Product_Master_Design.md
│   ├── Advocated_Data_Management_Standards.md
│   ├── Domain_Module_Design_Standard.md
│   ├── Semantic_Module_Design_Standard.md
│   ├── Search_Module_Design_Standard.md
│   ├── Prediction_Module_Design_Standard.md
│   ├── Observability_Module_Design_Standard.md
│   └── Memory_Module_Design_Standard.md
│
└── prompts/                 ← how to use the standards
    ├── Skill_Conversion_Prompt.md
    ├── Design_Data_Product_Starter.md
    └── Access_Data_Product_Starter.md

---

The Six Modules

Module	Purpose	Database pattern
Domain	Core business entities — source of truth	{Name}_Domain
Semantic	Metadata layer enabling agent discovery	{Name}_Semantic
Search	Vector embeddings and similarity search	{Name}_Search
Prediction	Feature store and ML prediction storage	{Name}_Prediction
Observability	Event tracking, quality monitoring, lineage	{Name}_Observability
Memory	Agent state, learning, and design memory	{Name}_Memory

Each module is independently deployable and composable. A data product can implement any combination.

Memory Module — Two Kinds of Memory

The Memory module hosts both:

• Runtime memory — agent sessions, interactions, learned strategies, user preferences, discovered patterns
• Design memory (Documentation Sub-Module) — architectural decisions, business glossary, query cookbook, change history

Both are temporal, append-oriented records of history. All documentation tables live in {Name}_Memory — the data product is fully self-contained with no external dependencies.

---

Deployment Order

Phase	Modules	Notes
1	Memory, Semantic	Memory hosts documentation tables needed by all modules; Semantic hosts discovery metadata needed by all modules — both must exist before any other module deploys
2	Domain, Observability	Domain is the entity foundation; Observability begins monitoring Domain immediately
3	Search, Prediction	Both require Domain entities to exist first

---

Prompts

Skill_Conversion_Prompt.md

Converts all design standard documents into a single unified Claude skill (ai-native-data-product.skill) with progressive disclosure:

• SKILL.md — always read by orchestrator and sub-agents; architecture, naming conventions, documentation capture protocol, routing instructions
• modules/{module}.md — read on demand; full DDL templates, design decisions, integration patterns, checklists for each module

Run this prompt whenever design standards are updated to regenerate the skill.

Design_Data_Product_Starter.md

Starting prompt for designing a new data product. Directs an orchestrator agent to load the unified skill and spin up sub-agents for each module.

Access_Data_Product_Starter.md

Starting prompt for an agent consuming an existing data product. Guides autonomous discovery via the Semantic module.

---

Key Principles

1. Design documents are the source of truth — skills are derived, never edited directly
2. Independently deployable modules — each module is self-contained with its own database and data model
3. Self-contained data products — no cross-product database dependencies
4. Zero data duplication — all modules join back to Domain via foreign keys; Teradata co-location makes this efficient
5. Agent-native design — queryable metadata, standard patterns, and multi-hop relationship discovery enable autonomous operation
6. Design memory — every module records its architectural decisions into the Memory module's documentation tables during the design process

---

Getting Started

Designing a new data product

1. Review AI_Native_Data_Product_Master_Design.md for architecture and naming conventions
2. Generate the unified skill using Skill_Conversion_Prompt.md (attach all design standards; run once per framework update)
3. Use Design_Data_Product_Starter.md to orchestrate the design

Updating the design standards

1. Edit the relevant design standard document(s)
2. Regenerate the unified skill using Skill_Conversion_Prompt.md with the updated document(s) and the current skill attached

---

Design Standard Versions

Document	Version
AI_Native_Data_Product_Master_Design	1.8
Domain_Module_Design_Standard	2.3
Semantic_Module_Design_Standard	2.5
Search_Module_Design_Standard	1.5
Prediction_Module_Design_Standard	1.6
Observability_Module_Design_Standard	1.4
Memory_Module_Design_Standard	1.6

---

License

Copyright © 2025-2026 Teradata Corporation

Licensed under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0). See LICENSE.md for full terms.

---

Acknowledgments

Developed by Teradata's Worldwide Data Architecture Team, Field Technology Organization.

---

Last Updated: March 2026
Status: Active Development — documentation-memory-merge branch