Cara Jade Catalano

I conduct observational research on the human voice — how it changes under autonomic intervention, how it carries signatures of trauma and recovery, and how motor-rhythm analysis can open communicative channels for individuals without functional speech. My work emphasizes long-duration naturalistic recording and rigorous statistical treatment of single-subject corpora.

Those commitments extend across a wider research program. I design and operate evaluation frameworks across 21 active project domains — rubric construction from first principles, cut-score derivation from empirical distributional analysis, inter-annotator agreement protocols applied to both human raters and LLM-as-rater pipelines, and longitudinal regression testing to surface capability drift over time. The same standards apply to my work on agentic AI systems: multi-turn behavioral consistency, failure-mode decomposition, distributional shift detection, and safety governance architecture with hard approval gates before any state-mutating action. Domains have included generative speech models, educational content quality, AI-generated publishing pipelines, citizen science classification corpora, and behavioral evaluation of non-verbal subjects — each requiring a measurement approach built from scratch, with no inherited rubric and no prior standard to anchor against.