Case studies

Open-source tools, applied research, and production-grade delivery outcomes.

Open Source / AI Security

OntoGuard: Production-Grade Ontology Firewall for AI Agents

Built and open-sourced a production-grade ontology firewall that validates AI agent actions against OWL-encoded business rules in real-time. Enforces role-based access control, amount constraints, time windows, and compliance rules. Provides human-readable explanations for every blocked action and a full audit trail for regulatory compliance.

19

GitHub stars

3

Enterprise domains demonstrated

5.2K

Views on launch article

OWL OntologySHACL ValidationPythonAI Governance
Open Source / Data Intelligence

Power BI Ontology Extractor: Unlocking 20M Hidden Ontologies

Built a tool that transforms Power BI data models into AI-ready ontologies — automatically extracting entities, relationships, and business rules from DAX measures. Generates OWL ontologies and MCP-compatible semantic contracts, enabling AI agents to reason over existing dashboard intelligence.

29

GitHub stars

25K

Views on companion article

20M+

Addressable Power BI dashboards globally

Power BIOWL OntologyMCP IntegrationPython
Open Source / Agentic Infrastructure

Self-Healing Multi-Agent System with MCP

Built an open-source multi-agent system that uses OWL ontologies and Model Context Protocol (MCP) to automatically adapt when database schemas change. Zero manual code fixes required on schema drift — the ontology layer detects changes and re-maps agent queries in real-time.

15

GitHub stars

6

Forks

5.4K

Views on companion article

MCPOWL OntologySelf-HealingMulti-Agent
Research / Benchmarking

AI Agent Backend Benchmark: PostgreSQL vs Vector DB vs OWL Ontology

Conducted a rigorous three-way benchmark testing PostgreSQL, vector databases (Pinecone), and OWL ontology as backends for AI agent decision-making. Tested across loan approval domain with regulatory constraints. Conclusion: none of them won alone — a query router combining all three is the production answer.

3

Backend approaches benchmarked

19

Files in companion GitHub project

1.1K

Views on companion article

PostgreSQLPineconeOWL OntologyBenchmark

Open source

Production-grade tools we build and maintain. Featured projects first.

More open-source work

foundry-ontology-open

Open-source OWL/SHACL export bridge from Palantir Foundry's proprietary ontology format. The vendor-neutral exit ramp for Foundry-locked ontologies.

4114Python
PalantirOWLSHACLFoundry

universal-agent-connector

MCP infrastructure with ontology-driven semantic routing enabling AI agents to connect to any enterprise data source.

75Python
MCPSemantic RoutingEnterprise

legacy-to-logic

Transforms legacy database schemas into AI-ready ontology infrastructure. The CLI tool behind the Extract-Model-Deploy methodology.

41Python
Legacy MigrationOntologyCLI Tool

ai-agent-backend-benchmark

Three-way benchmark: PostgreSQL vs vector DB vs OWL ontology as AI agent backends. Spoiler: none of them won.

40Python
BenchmarkPostgreSQLVector DBOntology

from-protege-to-production-python

End-to-end codebase for integrating OWL ontologies built in Protégé directly into Python-based AI pipelines.

112Python
OntologyPythonOWLProtégé

napkin-architect

Sketch-to-infrastructure-blueprint tool. Draw on a napkin, get production architecture. AI-powered technical design.

31Python
AIArchitectureDesign

agentic-mesh-security

SHACL-Gated A2A Router. Blocks cross-platform privilege escalation before it reaches the target agent.

40Python
SHACLA2ASecurity

ontology-rag-firewall

OWL/SHACL validation for unstructured data. Treats the LLM extractor as an untrusted agent. Catches what RAG pipelines miss.

50Python
OWLSHACLRAG

owl-portability-layer

Seven platform adapters — Google, Microsoft, Palantir, ServiceNow, Salesforce, AWS, IBM. One OWL/SHACL constraint layer across all of them.

255Python
OWLSHACLPortability