Summary
I build AI systems that enterprises can actually trust. After 20+ years in oil & gas, from field operations to data architecture, I've learned that in enterprise AI, accuracy is necessary but not sufficient. The harder challenges are evidence, accountability, and the judgment layer between AI output and human decisions.
Recent Projects (2024–2026)
- P&ID Extraction Engine: Computer vision extraction for engineering drawings in production for engineering clients; 491-sheet corpus, 54,470 detections, evidence viewer for every value
- KapwaCare: Multi-tenant home care operations platform, live since July 2026; 663 authorization gates, row-level security, Canadian payroll
- Harness Council: Multi-model coding-agent orchestration with gates, provenance, and measured outcomes; council beat any single agent in 13 of 14 recorded rounds
- Cartouche v2: Reliability-first document-control factory: agents at build time, deterministic conductor at run time
- Amphigraph: Evidence-first classification engine: 500K+ work orders with full provenance trails
- OpsWorth Doc: Engineering document intelligence: 39% to 93% accuracy on domain-specific ISA symbols
- ProductionAnalytics: SAGD production analytics, forecasting, and optimization
Experience
Self-Employed
AI Architect / Independent Consultant
January 2019 – Present · Calgary, AB
- Consulting for industrial clients: Suncor Energy, Pembina, Cenovus, Formosa Plastics, Imperial Oil, Cameron LNG, Chevron Phillips, ConocoPhillips, ExxonMobil, Williams, OPTA
- Built and delivered the P&ID Extraction Engine (2026): RF-DETR and SigLIP pipeline over a 491-sheet client drawing set, detector mAP 0.735, zero false positives on the held-out truth set; launched the P&ID Data Extraction service in July 2026
- Built and operate KapwaCare (2026): multi-tenant home care platform in production for an Alberta agency under the Health Information Act; scheduling, visit verification, payroll, and billing; 7,700+ automated tests
- Designed the Harness Council (2026): governed orchestration of Claude Code, Codex, and DeepSeek coding agents with 76 recorded decisions; used to build InqWeave and review three other codebases
- Author of The AI Capability Series (2026): seven articles on building AI capability in industrial organizations, published on LinkedIn and royabes.com
- Architected Amphigraph: 156-feature classification engine processing 500K+ maintenance work orders with full evidence trails
- Built multi-agent pipelines for industrial document intelligence using fine-tuned computer vision + LLM extraction, 93%+ accuracy on domain-specific ISA symbol detection
- Designed maintenance intelligence systems for work order classification, failure pattern analysis, and predictive maintenance pipelines
- Multiple production AI systems shipped to live industrial environments
Chevron Canada
Senior Production Analytics Specialist
January 2017 – December 2018
- Architected transition from Excel workflows to enterprise BI (Power BI, Spotfire), 80% reduction in analysis time
- Implemented ML models for production optimization, improving forecast accuracy by 35%
- Enabled precise production balancing delivering $500K+ monthly cost savings
- Enhanced SCADA and OSI PI data pipelines to 99.5% data reliability
- Awarded Employee of the Quarter for innovative analytics approaches
North West Redwater Partnership
Data Extraction & Quality Specialist
March 2015 – February 2016
- Extracted and cataloged instrument tag data from P&IDs with 99%+ accuracy
- Created systematic approaches reducing extraction time by 40%
Amec Foster Wheeler
Information Management & Document Control
October 2007 – March 2015 · staff, then independent contractor
- Directed migration of 100,000+ engineering documents to EDMS
- Implemented quality governance protocols achieving 99.8% compliance
- Designed SharePoint collaboration platforms serving 500+ users; reduced manual processing by 60%
Devon Energy
Information & Document Management Analyst
January 2012 – June 2012
- Built automated tools extracting equipment tags from engineering drawings
- Created structured repositories for 50,000+ technical documents
Shell Canada
Contractor: Web / Database Application Developer
January 2006 – September 2007
- Built enterprise web platforms serving 1,000+ users; automated transmittal system eliminating 20+ hours/week of manual work
- Only contractor to receive both "Contractor of the Month" and "Contractor of the Quarter"
Education & Certificates
- University of Illinois at Urbana-Champaign: Pattern Discovery in Data Mining (2015)
- Johns Hopkins University: Data Science Specialization Certificate (2015)
- Southern Alberta Institute of Technology: Records and Information Management (2008)
- St Joseph's College: Bachelor of Arts
Tech Stack
Python · PyTorch · LangChain/LangGraph · Anthropic & OpenAI APIs · HuggingFace · LoRA/QLoRA/PEFT · RF-DETR · SigLIP · YOLO · Neo4j · PostgreSQL · pgvector · Soufflé (Datalog) · FastAPI · vLLM · React · TypeScript · Next.js · Docker · GitHub Actions · GCP Vertex AI
Open to consulting engagements and senior AI roles.