Patrick Glatz

AI Workflow Design & Learning Systems

Designing governed workflows that preserve human judgment under pressure

glatz.p@gmail.com LinkedIn →

Summary

AI workflow designer specializing in constraint-based systems. Designs governed interaction models and decision systems that prevent task drift and preserve human judgment under constraint. Applies behavioral systems thinking to structure human–AI interaction under real-world constraints.

Systems

Governed Nonfiction Writing Companion

Designed a constrained AI-assisted writing system to preserve student authorship under AI assistance.

  • - Enforces explicit intent declaration before interaction
  • - Prevents automated idea generation and rewriting
  • - Blocks stage advancement without explicit user acknowledgment
  • - Structures fixed decision checkpoints
  • - Enforces an auditable record of decisions and user responsibility

Outcome: Prevents silent delegation and unauthorized authorship shifts.

Preventing Silent Delegation in Image Generation

Designed a structured intake system that prevents ambiguous delegation in generative workflows.

  • - Enforces explicit intent definition before generation
  • - Blocks execution until key decisions are resolved
  • - Surfaces implicit assumptions to block silent task completion
  • - Enforces structured representations of user intent prior to output

Outcome: Prevents implicit assumption filling under ambiguous instruction.

Wolverine Survival Simulation (Constraint-Based System)

Designed a governed simulation enforcing ecological constraint through irreversible decision-making.

  • - Enforces single-path, non-reversible decisions
  • - Constrains outcomes via rigid resource limits and environmental chance
  • - Prevents optimization strategies that bypass constraint
  • - Blocks adaptive rescue and instructor intervention during execution
  • - Enforces visible system state attributable to structure

Outcome: Prevents false signals of learning driven by unstructured effort.

Advisory Drift in Adaptive Systems (LLM Study)

Designed a test framework to expose how AI systems revise decisions when new context is introduced mid-process.

  • - Exposes decision variance between staged and baseline conditions
  • - Surfaces patterns of AI refinement and replacement
  • - Exposes inconsistent problem reconstruction
  • - Exposes over-weighting of visible constraints at the expense of critical ones

Outcome: Prevents unnoticed problem reconstruction during decision-making.

Professional Experience

Special Education Teacher – Autism Program

Somerdale Park School District | 2024–Present

Designed and operated structured instructional and decision systems in high-constraint classroom environments.

  • - Built AI-assisted planning workflows with enforced constraints
  • - Structured stepwise input processes
  • - Designed reinforcement systems to reduce reliance on real-time judgment
  • - Enforced explicitly visible system states through structured environments
  • - Enforced system adherence across staff to maintain consistent decision-making

Behavior Support Teacher (K–6)

Waterford Township School District | 2022–2024

Designed behavior-support systems under real-time instructional constraints.

  • - Developed workflows defining explicit response conditions
  • - Structured behavioral data into actionable systems
  • - Constrained decision paths through clear, predefined rules
  • - Reduced variability in instructional decisions across multiple operators

Special Education Teacher – Multi-Disability Program

Somerdale Park School District | 2016–2022

Designed durable instructional and behavioral systems for complex environments.

  • - Built structured routines to ensure stability across staff turnover
  • - Enforced reinforcement systems aligned strictly to observable conditions
  • - Enforced repeatable, constrained workflows to reduce cognitive load
  • - Maintained system integrity under real-time classroom pressure

General Education Teacher

St. Raphael School | 2013–2016

Designed structured instructional systems aligned to defined learning outcomes.

  • - Built scaffolded lesson sequences with explicit stage progression
  • - Structured curriculum into repeatable instructional workflows
  • - Enforced structural consistency to reduce instructional variance

Education

B.A. in Education, Rowan University — Magna Cum Laude

B.A. in American Studies, Rowan University — Magna Cum Laude

  • NJ Standard Elementary Teacher & Teacher of Students with Disabilities Certification
  • CPI Nonviolent Crisis Intervention Training
  • Applied Behavior Analysis (5+ years practical experience)