William Craig

ML Engineer | Safety Systems

William Craig is applying lightweight computer vision techniques to real automotive safety problems.
His work is inspired by a 6-year career in automotive research and accident investigation.

Core Competencies

ML Engineering

  • Computer Vision & Perception
  • Supervised & Unsupervised Learning
  • Real-time Inference & Deployment
  • Edge Case & Robustness Testing

Safety-Critical Systems

  • ADAS/ADS Validation (CARLA/AutoSAR)
  • Safety Risk Management (6+ years)
  • Incident Investigation (300+ cases)
  • Hazard Analysis & FMEA

Leadership & Standards

  • Technical Committee (NATM)
  • Cross-Functional Consensus Building
  • Production System Deployment
  • Technical Witness & Communication

About Me

I build ML systems for autonomous vehicles and safety-critical applications. My work spans computer vision for real-time perception, unsupervised learning for robust deployment, and developing safety validation frameworks for AI systems.

What makes my approach different: I spent 6+ years investigating what happens when systems fail. I've analyzed 300+ vehicle accidents, designed safety-critical electronics that shipped to production, and served as a technical witness explaining failures to courts and stakeholders. That background means when I build ML systems, I'm already thinking about edge cases, failure modes, and safety validation—not as an afterthought, but as part of the design.

I've led teams (Formula SAE captain), gained multi-stakeholder consensus on critical safety issues, and served on technical standards committees (NATM). I understand both the engineering and the organizational challenges of deploying AI safely.

Experience

ML for Safety-Critical Systems

Developing computer vision systems for autonomous vehicle validation (CARLA/AutoSAR), fleet risk assessment, and real-time safety monitoring. Focus on robust deployment: unsupervised learning for diverse conditions, edge case handling, and failure mode analysis.

Vehicle Safety Engineering (6+ years)

Investigated 300+ incidents as corporate technical witness. Designed safety-critical electronics (2 patents). Found critical supplier documentation errors and coordinated multi-business-unit response. NATM technical committee (Member of the Year, 2022).

Systems Engineering Leadership

Led Formula SAE team through complete vehicle development. Built products from prototype to production. PE license in mechanical engineering. Track record of cross-functional leadership and stakeholder consensus-building.

Technical Expertise

  • ML Engineering - Computer vision, supervised/unsupervised learning, real-time inference, model deployment
  • Safety-Critical AI - ADAS validation, safety case development, edge case testing, failure mode analysis
  • Automotive & Robotics - ROS2, CARLA/AutoSAR, embedded vision systems, sensor integration
  • Safety Risk Management - 6+ years incident investigation, hazard analysis, technical witness, standards committees
  • Production Systems - Patent holder (vehicle electronics), shipped products, PE license

Education & Credentials

  • M.S. Engineering - Arizona State University (GPA: 3.76)
  • B.S. Engineering (Electrical Systems) - Arizona State University
  • Professional Engineer (PE) - Mechanical: Machine Design & Materials
  • Certified Fire and Explosion Investigator (CFEI) - NAFI
  • ASE Certified Master Light Automotive Technician

Blog

Coming Soon: First Post

I'll be sharing insights on ML system design, autonomous vehicles, and bridging the gap between algorithms and hardware.

Read More →

Unsupervised Clustering of Vehicle Lights: Lessons Learned

Practical insights from integrating machine learning models into a smart trailer for real-time control.

Read More →

Technical Notes

📝

ML System Design Patterns

Collection of architectural patterns for deploying ML models in production environments.

MLOps Architecture
🤖

ROS2 Quick Reference

Personal reference guide for ROS2 development, node architecture, and common patterns.

ROS2 Robotics
🚗

Automotive CAN Protocol

Notes on CAN bus analysis, J1939 protocol, and automotive diagnostics for ML feature engineering.

Automotive Protocols

PyTorch Performance Tips

Optimization techniques for training and inference, memory management, and profiling.

PyTorch Optimization
🔧

Hardware-ML Integration

Practical considerations for deploying ML on embedded systems and edge devices.

Embedded Edge AI
📊

Time-Series ML Methods

Approaches for time-series forecasting, anomaly detection, and sequential data modeling.

Time-Series ML