Native semantic models
Snowflake Semantic Views and Unity Catalog metric views place measures, dimensions and relationships near their execution and governance platform.
CONTEXT & SEMANTIC LAYERS / HANDS-ON ACCELERATORS + SOLUTIONING
Business meaning.
Built into the system.
I build reusable semantic and context accelerators and lead the capability showcases, platform demos and proposals that turn them into a customer solution. My work spans Snowflake, Databricks and GCP, with the representation chosen for the use case and scale.
Snowflake Semantic ViewsUnity Catalog metric viewsKnowledge graphs / Neo4jMCP tool interfacesSpanner Graph prototypesOKF / Apache Ossie
Original work: launched a Context & Semantic Layer
offering and designed an enterprise architecture across Finance,
Merchandising, Supply Chain and Store Operations. Offering snapshot:
four pursuits, approximately $1.4M pipeline and two late-stage
negotiations. This is an earlier offering-specific snapshot, not an
amount to add to the approximately $7.2M current GTM team pipeline.
THE CAPABILITY
Define the measure.
Connect the business around it.
01
Semantic layer
Business entities, grain, measures, dimensions and joins become reusable contracts. BI, SQL and agent tools use a consistent definition with explicit versions and aggregation rules.
NATIVE METRICS / SYNONYMS / JOINS / VERSIONING
02
Context layer
Connect those definitions to the entities, transformations, sources, owners and policies that explain them. Give an agent the relevant, authorized relationships and provenance for its task, including cross-domain questions that require customer and product knowledge together.
BUSINESS RELATIONSHIPS / DERIVATION / POLICIES / EVIDENCEBUILT PROTOTYPE / MEMBERSHIP RETAIL / GCP
Two domains.
One deeper business question.
“Which customer demographics over-index on our strong renewal drivers in pantry?”
I built a Spanner Graph context-layer prototype connecting Merchandising and Customer domains. The question needs product/category knowledge, purchase relationships, demographic cohorts and a shared definition of renewal-associated products.
Personal build ownership2 connected domainsSpanner GraphCross-domain reasoning
Original prototype confirmed by me. The entity schema, serving controls and numerical walkthrough below are independent portfolio illustrations. Client data, schema, code and measured outcomes are not reproduced.
SYNTHETIC ANALYSIS / EXPLICIT DENOMINATORS
Make “over-index” inspectable.
Distinct buyers of an approved pantry product set ÷ eligible members in each cohort. Divide that penetration by the same-window population baseline.
Illustrative window: Jul–Sep 2026 · minimum cohort: 100 members · mutually exclusive age bands · population includes buyers and non-buyers| Cohort | Buyers / members | Penetration | Index |
|---|
Association, with a defined population. “Renewal
drivers” refers to an approved renewal-associated feature or product
set. This comparison does not establish causal lift. Count members
once per window, include non-buyers in denominators, avoid
purchase-edge fan-out, and suppress small cohorts. The illustration
demonstrates buyer penetration; renewal prediction or causal
estimation requires a separate validated analysis.
END-TO-END REFERENCE DESIGN
Follow the definition,
the calculation and the context.
Governed SQL or GQL execution supplies numerical evidence. The graph connects business relationships and provenance; validated tools bind definitions, scope and evidence to the consumer.
Portfolio reconstruction of the capability. Service bindings are reference choices; this browser does not connect to a cloud account, Neo4j or a live MCP server.
ILLUSTRATIVE CONTRACT WALKTHROUGH / SYNTHETIC RECORDS
An answer you can
inspect all the way back.
Choose a business question and inspect the metric definition, context relationships, tool request and evidence. Calculations below run locally against small synthetic fixtures.
SYNTHETIC RESULT
Definition v1.0 · fixed as-of time · bounded scope
01 / DEFINITIONSelect the metric contractGrain, aggregation, units and allowed dimensions.
02 / CONTEXTRetrieve the relevant relationshipsSources, ownership and policy provenance.
03 / EXECUTIONCall the bounded metric toolTrusted identity and access enforced by services.
04 / EVIDENCEReturn an inspectable resultValue, definition version, scope and as-of markers.
Metric definition
Context relationships
Illustrative MCP tool request
Answer evidence
REPRESENTATION CHOICES / PLATFORM + PORTABILITY
Choose the right form
for meaning and context.
I’ve built showcases and demos using native semantics, knowledge graphs and portable representations, selecting the approach for the use case and scale.
Apache Ossie
Formerly Open Semantic Interchange (OSI): JSON/YAML models for exchanging dataset, relationship and metric definitions. Incubating; the core specification is evolving.
Read the draft specification ↗Open Knowledge Format
OKF organizes context in Markdown with YAML frontmatter and linked concepts. Use it for definitions, curated knowledge and provenance; choose a separate serving/query runtime.
Read the OKF specification ↗Knowledge graphs
Neo4j or Spanner Graph represents domain entities and relationships for traversal and structured reasoning. The Spanner prototype shows the two-domain application.
Follow the membership example ↑Interchange requires explicit mappings and tests for grain, cardinality, aggregation, versions and vendor extensions. The JSON downloads on this page are illustrative contracts, not Ossie-conformance claims or automatic platform converters.
ARCHITECTURE JUDGMENT
The choices behind
a useful context layer.
Native or portable?
Use platform-native semantics when execution and governance should stay close to the data. Translate shared business contracts deliberately across representations, with tests for grain, joins and aggregation.
Which context belongs in a graph?
Model business relationships and derivation paths that answer a real question. Keep sources, effective dates and versions. Structured graph queries can also produce aggregates. Reconcile them with the same population, grain and business definitions.
What can the agent access?
Enforce data grants, graph audience filters and tool permissions independently. Bound queries, traversals and result sizes. Reconcile freshness and versions before composing evidence.
CONNECTED TO MY PROJECT WORK
Three places where
meaning changes the outcome.
01 / METRICS + CONTEXT
Beauty data platform
Shared cashflow/RGM measures, financial grain, scoped SQL tools and authorized document evidence.
Inspect the walkthrough ↗02 / DECISION CONTEXTRetail recommendations
Vehicle requirements, retailer scope, stock freshness and explainable product alternatives.
Inspect the walkthrough ↗03 / ENTITY MEANINGGlobal product master
Regional taxonomies, product identity, domain constraints, source crosswalks and consistent publication.
Inspect the walkthrough ↗PLATFORM REFERENCES