Most applied-AI engineers can't get into the cleared room. Most cleared DoD veterans can't build the
pipeline. I do both — 15+ years of US Navy, USMC, and federal domain depth paired with
hands-on, production applied-AI work. Active Top Secret clearance. Navy veteran.
Right now I build production AI pipelines at an AEC-focused startup — local LLMs, RAG semantic search over
pgvector, n8n orchestration, and containerized infrastructure (Docker, PostgreSQL) — proof the skills are
current and production-grade. I build for the environment defense actually runs in: on-prem and air-gapped,
least-privilege, no dependence on commercial cloud. What I really do is translate between the mission and
the machine — elicit requirements from government stakeholders under real constraints, then build the
system that meets them.
Microsoft 365 • SharePoint • Teams • Graph API • OAuth2 / Azure AD
Power BI (DAX, M) • SQL Server / T-SQL • Excel VBA • Requirements Analysis
Projects
Production AI Competitive Intelligence Pipeline
AI Pipeline & Automation Engineer
End-to-end competitive intelligence pipeline for an AEC-focused startup — ingesting RSS, YouTube,
and podcast content, processing 500+ documents weekly through a local LLM
(Ollama/Llama 3.1) for relevance scoring, entity extraction, and semantic clustering.
Delivered automatically to SharePoint and Microsoft Teams via Graph API.
500+
15+
0
Autyvia AEC Intelligence Pipeline
Designed and deployed production pipeline ingesting RSS, YouTube, and podcast sources through Ollama/Llama 3.1 for relevance scoring and entity extraction
Built semantic clustering pipeline using pgvector embeddings to automatically group content by topic, eliminating a flood source diluting content quality
Architected 15+ n8n automation workflows spanning ingestion, LLM processing, deduplication, and automated delivery to SharePoint and Microsoft Teams via Graph API
Designed config-driven Python scraper pipeline to expand sourcing beyond RSS feeds
Migrated production pipeline from MongoDB to PostgreSQL — schema design, backfill workflows, dual-write transition with zero data loss
Designed LLM input sanitization layer and least-privilege OAuth2 credential architecture across Microsoft 365 integrations
Across a federal modernization program's master set of ~30,000 requirements, identified
~4,000 that had slipped through unmatched to any test case, then built a Python/NLP pipeline
(Jaccard similarity, clustering, context-aware topic extraction) to match them — cutting that validation
from 3–4 months to < 2 hours (≈ 98% reduction).
Built Sept–Nov 2025.
~4,000
98%
60→85%
Intelligent Requirements-to-Test Case Matching System
Worked from an NP2 master requirements set of ~30,000; targeted the ~4,000 with no test case on record
Jaccard-based similarity scoring with verb routing and hierarchical clustering
Context-aware topic extraction to pull meaning from dense technical language
Adaptive domain acronym harvesting (633 Navy terms, 98% extraction accuracy)
Raised requirement-to-test-case traceability confidence from 60–70% to 75–85%
Cut analyst review time ~50% per requirement; cross-product transfer saved 40+ hours per product family
Built on Controlled Unclassified Information (CUI) — results described here; source data not disclosed.
PythonNLPMLText AnalyticsJaccardData Engineering
Executive Metrics Dashboard with Advanced Analytics
BI Developer & Data Analyst
Executive dashboards with quad-chart slicers and 15 KPIs. Integrated data from
13 formats via Power Query ETL, increasing reporting metrics by 496% and
visual elements by 389%.
87%
94%
98.2%
Power BI Performance Management Dashboard
Interactive quad-chart dashboard with drill-down capabilities
DAX measures for complex metric calculations
Automated ETL aggregating 13+ data source types
496% increase in functional test metrics; 389% more visual elements
Clean, normalized data models ensuring accuracy and integrity
Links: Government client under NDA.
Power BIDAXPower Query (M)SQL ServerETL
High-Speed Requirements Validation Macro
VBA Developer
Macro validates 1,592 requirements across 471 test cases in
< 1 minute vs 7+ hours manual. RTVM identified a
3.4% coverage gap (62/1,806), enabling 100% verifiable coverage.
<1 min
98.6%
RTVM
Excel VBA Requirements Traceability Automation
Validates 1,592 requirements vs 471 test cases in <1 minute
RTVM revealed 3.4% coverage gap (62 of 1,806) → enabled 100% coverage
Reusable macro framework for ongoing validation cycles
Automated gap detection & quality checks
Links: Federal contract (sensitive data).
Excel VBAAutomationQAData Validation
Enterprise Budget Consolidation System
Database Developer & Solution Architect
Consolidated $628M across 600+ accounts from siloed Excel files into a unified
SQL Server / MS Access system. Regex-driven parsing & VBA procedures enabled rapid reconciliation and
executive visibility, reducing manual work by 75%.
$628M
600+
75%
SQL Server / MS Access Financial Management Database
Consolidated $628M across 600+ accounts into a unified platform
MS Access front-end with SharePoint 2010 back-end providing real-time tracking and query reporting for
170M+ data points across depots. VBA ETL imports heterogeneous Excel feeds, automates discrepancy analysis,
and accelerates processing by 65%.
170M+
24/7
65%
SharePoint-Integrated Tactical Vehicle Tracking System
Real-time tracking for 170M+ data points across depots
VBA-driven ETL importing from multiple disparate sources
Dynamic discrepancy identification and resolution workflows