Twenty years of high-pressure simulation. Now applied to AI reasoning and agent evaluation.
Twenty-plus years in high-stakes simulation for clinical medicine, emergency response, police training, and legal evaluation. Environments where failure is instructive, ambiguity is deliberate, and edge cases are the point.
Since 2025, I have applied that background in high-risk simulation across health, emergency response, and legal domains to honing the methodology for AI systems: writing adversarial prompts, constructing complex evaluation scenarios and testing model reasoning across business, operational, and creative domains.
Strong writers make strong task creators. I am a professionally trained voice artist and producer with thirty-plus stage productions and ten-plus film credits. I know how personas behave under pressure.
Over two decades in high-stakes simulation across medical schools, police agencies, airport tarmacs, and oil rigs. The work ranged from scripted frameworks and self-directed character development to live improvisation across large-scale crisis scenarios. Layered personas, competing priorities, deliberate information gaps: the conditions that reliably surface reasoning failures.
Specialized in constructing tasks that surface model blind spots: ambiguous instructions, contradictory context, plausible-but-wrong information, and constraint conflicts. Experienced in structured rubric-based evaluation across safety, factual reliability, and instruction adherence.
Career-long background as a producer, festival co-founder, and production manager. Comfortable designing scenarios with multiple principals, conflicting priorities, and the kind of organizational ambiguity real agent tasks have to navigate.
Formally trained voice artist and producer. I write personas that hold up: consistent tone, realistic communication patterns, authentic professional register. This translates directly to crafting believable email threads, team messaging, and calendar contexts that test an agent's ability to read human intent.
Experienced with structured data entry and JSON-based task definition from AI evaluation work. Prior corporate operations background provides comfort with workflow tools, rubric frameworks, and systematic documentation under volume.
Scenario work spans business operations, healthcare-adjacent simulation, emergency response, legal-adjacent evaluation, e-commerce, and creative storytelling. Wide domain coverage produces richer, less predictable task sets.
Architect a scenario once, then port it across regulated, commercial, and creative domains. The underlying pressure structures (escalation, ambiguity, divided attention) transfer cleanly; only the surface vocabulary changes. This is what twenty years of cross-industry simulation work makes possible: a clinical de-escalation scenario re-skinned for customer service, a legal evaluation rebuilt for e-commerce dispute handling, an emergency triage adapted to AI agent escalation paths. Same evaluative spine, new domain.
Conduct structured evaluation of LLM outputs against rubric-based safety, alignment, factual reliability, instruction-following, and performance criteria. Design complex scenario-based evaluations across business, operational, healthcare-adjacent, e-commerce, and creative domains. Experienced with long-context reasoning, agent behaviour evaluation, adversarial prompt design, and concise analytical feedback.
Delivered realistic patient and subject simulations for medical, clinical, and institutional training programs across 18+ years, and simulated legal proceedings across family, criminal, and civil law. Worked from collaboratively developed scripts to build layered personas, maintained role fidelity under adversarial conditions, and provided structured post-scenario debriefs. Extended to police conflict management simulation and emergency response exercises across airport tarmacs, terminals, and oil rigs with first responders and inter-agency coordination.
Founded and operate a curated film festival, handling programming, vendor coordination, stakeholder communications, scheduling, and logistics across multiple concurrent priorities. Directly applicable to designing realistic multi-stakeholder professional environments for agent scenario tasks.
Production management for film and TV including Deafening Darkness (all-Deaf cast, ASL), Blood Riders, Better People, and WarJunk TV (History Channel / Discovery Network). Two decades of multi-location, multi-stakeholder production coordination; this is the organizational complexity realistic agent scenarios require.
Retail and e-commerce operator covering sourcing, pricing strategy, customer behaviour analysis, margin decision-making, and digital marketing. Provides grounded commercial reasoning for business-domain scenario design, including purchasing decisions, vendor negotiation, and conversion tradeoffs.
Two-year intensive in screen and stage performance, character construction, voice, and on-set protocol. The foundation of how I write personas that hold up under pressure: consistent tone, realistic communication patterns, and authentic professional register across long-form scenarios.
Institutional reference confirming clinical simulation work at Memorial University's Faculty of Medicine, including patient persona design and structured evaluation support.
Reference confirming adversarial simulation work for law enforcement conflict management training; response protocols tested under realistic pressure.
Reference confirming simulation support for airport emergency exercises involving tarmac and terminal scenarios, first responders, and inter-agency coordination.
Track completion confirming structured understanding of AI applications in business contexts, directly applicable to commercial domain scenario design.
Course completion in generative AI applied to business decision support, informing scenario realism for AI agent evaluation tasks in organizational settings.