Occams Consulting · Client: Keyera · Calgary, AB · Aug 2026 – Present

• • Build Databricks Apps and databases
• • Design data models and pipelines on the Databricks platform

Plains Midstream Canada · Calgary, AB · 2024 – May 2026

• • Manage integrity dig programs across multiple pipeline systems, coordinating with operations, engineering, and regulatory teams
• • Develop Power BI dashboards and reporting systems to consolidate pipeline integrity data from multiple sources
• • Ensure regulatory compliance across all integrity activities and vendor operations

Plains Midstream Canada · Olds, AB · 2023 – 2024

• • Managed three of six Control Center consoles overseeing real-time pipeline monitoring and operations
• • Led team of 18+ personnel ensuring 24/7/365 coverage for alarm response and incident management
• • Coordinated emergency response activities and incident investigations across multiple pipeline systems

Plains Midstream Canada · Olds, AB · 2019 – 2023

• • Led Alarm Management program ensuring optimal system health, sensitivity, and reliability
• • Oversaw Compliance function maintaining regulatory requirements across control center operations
• • Managed Training program ensuring controllers maintained proficiency in leak detection response
• • Administered Management of Change processes, completing thousands of MOCs without overdue items

Plains Midstream Canada · Olds, AB · 2018 – 2019

• • Served as technical SME for SCADA systems, leak detection, and control center processes
• • Supported continuous improvement initiatives for leak detection and alarm management systems

Plains Midstream Canada · Olds, AB · 2012 – 2018

• • First Shift Supervisor in Olds Control Center, establishing on-shift supervision framework
• • Provided real-time oversight of leak detection alarm response and analysis
• • Mentored controllers on pipeline hydraulics, leak detection interpretation, and alarm response

Plains Midstream Canada · Lloydminster & Olds, AB · 2003 – 2012

• • Monitored real-time pipeline data using SCADA systems to detect irregularities and respond to alarms
• • Operated SSPL and Manitou consoles, managing pipeline hydraulics across multiple systems
• • Provided on-shift leak detection alarm response and analysis for pipeline operations

Industrial pipe inspection system detecting metal loss and welds via multi-method computer vision.

• • ML Models: YOLO + Faster R-CNN object detection with LLM vision validation for ensemble consensus
• • Pipeline: CLAHE normalization → Gaussian background removal → percentile thresholding → morphological dilation → connected components with 8 statistical filters
• • Inference: 640×640 sliding window tiling with 25% overlap → ONNX Runtime (GPU/CPU) → NMS (IoU 0.5)
• • Fusion: IoU-based transitive grouping via Union-Find, multi-method consensus merging, confidence-weighted NMS
• • Scale: Panoramic HAL sensor images up to 9999×1024px; Traditional: 100-300ms, YOLO: 50-150ms (GPU)

Personal, single-user asset integrity workflow system for tracking pipeline dig programs, schedules, vendors, documents, financials, DWRs, and weekly reporting.

• • Operational Model: GWD/girth-weld dig records, CWPs, vendors, schedule states, checklists, scouting, ILI context, notes, and workflow steps
• • Data Workflows: CSV/Excel import/export, DWR import, conflict handling, Supabase/PostgreSQL persistence, RLS, storage, and SSR authentication
• • Document Intelligence: Claude extracts structured fields and GWD/CWP/vendor links from uploaded emails and documents at ingest; PostgreSQL full-text search (TSVECTOR/GIN, ts_rank) tuned for operational terminology
• • Agentic Chat: 11-tool assistant that picks its own retrieval strategy—document search, dig filtering, vendor summaries, and schedule queries—over operational data
• • Reporting: Weekly report generation, Gantt/schedule views, open dig tables, financial reconciliation, PO/vendor ticket tracking, and PDF/export flows

Autonomous ground robot using computer vision for real-time object tracking, line following, and gesture control.

• • Hardware: Raspberry Pi running vision and planning, ESP32 motor controller over UART, USB/CSI/OAK-D depth cameras, and optional LiDAR for obstacle mapping
• • Vision: OpenCV, MobileNet-SSD object detection, and MediaPipe hand/pose/face tracking for object following and gesture control
• • Control: State machine across 9 vision modes, PID-style gimbal tracking, and HSV-based line following

Orchestrates 20 specialized LLM agents to autonomously manage software projects through hierarchical task delegation.

• • Multi-Provider: Routes tasks across OpenAI, Anthropic, and Google models based on agent specialization and capability scores
• • Hierarchy: 4 supervisor agents analyze complexity, identify skills via pattern matching, delegate to 16 specialists using capability-based routing
• • Coordination: Event-driven handoff (1:1) and broadcast (1:N) patterns with acknowledgment tracking and timeout handling via EventBus/MessageQueue
• • Resilience: Token-aware rate limiting, round-robin API key rotation, semantic response caching (0.92 similarity), automatic provider fallback
• • Autonomy: Claude Agent SDK integration for 30+ hour sessions with checkpoint/rewind, subagent spawning, and configurable permissions

Agents trained from scratch to play Super Mario Bros, comparing evolutionary and gradient-based learning on the same control problem.

• • Methods: NEAT neuroevolution vs. PPO deep reinforcement learning
• • Training Design: Engineered the reward shaping and structured game-state inputs the agents learn from