AI SAFETY · RED TEAMING · MODEL EVALUATION
Caroline V. James
AI red teamer and model evaluator focused on adversarial testing, safety, and model behavior.
SELECTED WORK
ABOUT
My work focuses on how generative AI behaves in difficult, ambiguous, and adversarial contexts. I design and evaluate prompts across safety domains, test indirect prompt injection and other attack strategies, and analyze model failures involving groundedness, bias, and reasoning.
EXPERIENCE
My work spans adversarial prompt design, red teaming, indirect prompt injection, model evaluation, and safety-focused dataset development. I design realistic personas, dialogues, and use cases to probe how models behave under ambiguity, social pressure, hidden instructions, and competing objectives. My background in qualitative research and the social sciences informs an approach focused not only on whether a model fails, but on the mechanism that makes the failure possible.
Adversarial prompt design & red teaming · Indirect prompt injection / XPIA · Model evaluation & safety analysis · Synthetic data & dialogue · Persona & use-case development · Qualitative research
CONTACT
Caroline V. James
Available for AI safety, red-teaming, and model-evaluation work.