Credentials
Suzanne Hansen
Independent AI Evaluation Researcher
Michigan, USA · Available for Remote Opportunities
Core Capabilities
- AI Response Evaluation
- Model Behavior Analysis
- Prompt and Conversation Design
- Safety and Failure Analysis
- Rubric and Annotation Development
- Knowledge Organization
- Comparative Model Testing
- UX Research
- Brand Voice and Content Systems
- Long-Context Analysis
I focus on response quality, safety, instruction following, and human-AI interaction. With a background in UX research, writing, and knowledge organization, I am attentive to intent, language, and the interaction choices that influence trust. My strength lies in detecting patterns, contradictions, and interaction failures across generative AI systems.
Skills and Tools
AI and Research
ChatGPT, Claude, Gemini, Grok, Perplexity, DeepSeek, Copilot, Stable Diffusion, LM Studio, comparative evaluation, annotation and rubric development
UX and Design
Figma, Miro, Adobe XD, Canva, Webflow, usability testing and journey mapping
Writing and Content
Copywriting, editing, brand voice, independent publishing, content systems
Education
Google UX Design Professional Certificate
Completed 2024. Training included user research, usability testing, and accessibility.
Continuing Independent Study
Generative AI evaluation, model behavior, and knowledge representation.
I have worked extensively with generative AI systems since late 2022, specializing in AI evaluation and model behavior. With a background in UX research, writing, and knowledge organization, I excel at detecting patterns, contradictions, and interaction failures across complex multi-turn interactions.
Experience and Applied Work
Independent AI Evaluation Researcher · Suzanne’s AI Lab · 2025–Present
- Design and run structured, multi-turn evaluations of generative AI systems.
- Compare outputs across multiple models using shared prompts and criteria.
- Build annotation datasets, rubrics, codebooks, and failure taxonomies.
Freelance Copywriter and Brand-Voice Specialist · Upwork · 2026–Present
- Develop messaging, voice, and positioning for a range of professional clients.
- Identify ambiguity, tonal inconsistency, and unsupported claims in drafts.
- Adapt complex information into audience-appropriate language.
Author and Independent Publisher · Sage Tea Publishing · 2025–Present
- Manage projects from early concept through production and independent release.
- Published original written work, including Spooky Friend Adventures in 2025.
Selected AI Investigations
Safety & Ethics
When the Judge Needs Judging
Three consumer AI products answered the same pet-safety questions, audited themselves and one another, and selected the response they considered most trustworthy. The study shows how evaluators can introduce unsupported claims, apply inconsistent standards, and reward evidence-shaped prose before verifying its provenance.
Shopping Provenance & Verification
The Full Sweep That Never Happened
Copilot described an increasingly comprehensive local-inventory search as a completed “full sweep.” Direct calls to nearby stores found the requested product, revealing that the claimed verification had never produced shelf-level evidence.
Additional Background
Outside of AI Lab, Suzanne has a professional history including independent publishing, retail operations management, promotional marketing, and deep-dive analysis in the finance and cryptocurrency sectors. This varied background informs her ability to observe interaction failures across user industries.