About
Hi! I’m Matt Hartigan, Annotation Systems Architect. I focus on the operating systems behind AI data production: how annotation, review, evaluation, and feedback workflows transform expert decisions into defensible ground truth and effective model supervision. My work sits at the intersection of annotation governance, data-centric AI, model evaluation, human-in-the-loop systems, data quality strategy, AI infrastructure, and the emerging discipline of supervision system design.
My background spans product operations, AI data operations, annotation systems, model evaluation, and cross-functional delivery across autonomous vehicles, generative AI, NLP, computer vision, and regulated financial services. I developed my annotation systems architecture practice after building the workflows, metrics, roadmaps, dashboards, vendor models, and operating structures that turn ambiguous AI data needs into production-ready systems.
I help AI and data teams diagnose where annotation, review, evaluation, and feedback workflows lose signal. I translate that diagnosis into practical operating improvements across ground truth formation, quality metrics, benchmarking, workload planning, sampling, training, readiness, routing, escalation, and forecasting — so supervision systems become more explainable, useful, and defensible.
Research Framework
My consulting work draws from Governed Annotation Systems, an unpublished research framework I am developing around interpretive AI data work and data-centric AI. The framework examines how institutions produce, measure, review, route, train, and forecast annotation systems when quality depends on judgment rather than labels alone. The central thesis is public: AI data production requires stronger systems for turning expert decisions into defensible model supervision. Deeper materials remain reserved for consulting, advisory, partnership, or appropriate review discussions.