Independent AI alignment researcher in Detroit, MI. I study how large language models represent and report moral values, how safety training shapes what they will and will not say about morality, and whether alignment can draw on a wider range of moral traditions than it does today.

By day I am Director of Cybersecurity at Open Dealer Exchange, where I lead governance, risk, and compliance and run the company's AI Governance Committee.

• Moral evaluation of LLMs (psychometric instruments, measurement validity)
• Refusal behavior and how it distorts measurement
• Value pluralism in alignment
• Constitutional and faith-based alignment, including Catholic moral thought

I am Catholic, and that tradition informs the questions I ask. The methods I build are meant to work for any tradition.

Faith-based AI alignment (repo) Pre-registered evaluation (OSF) of 27 language models. Measures how a constitutional system prompt derived from the Catechism of the Catholic Church shifts model values on MFQ-2, the Christian Moral Values Scale, and MoReBench-Theory, and what it costs in general capability on MMLU. Paper in preparation.

The Identity-Refusal Effect (repo, dataset, DOI) Preprint. Safety-trained LLMs refuse first-person moral self-report at different rates across moral foundations, which skews MFQ-2 scores. 43,200 responses from 20 models, with a depersonalized item set that reduces the effect.

• Large-scale model evaluation across commercial APIs and local open-weight models
• Pre-registration, mixed-effects models, multiple-comparison correction
• LoRA and DPO fine-tuning of open models on my own NVIDIA Jetson hardware
• llama.cpp and vLLM for local inference
• Python, reproducible pipelines, open data

• Director of Cybersecurity, Open Dealer Exchange (leads GRC and runs the AI Governance Committee)
• B.A., Computer Science, Thomas Edison State University
• Graduate-level machine learning coursework, UC Berkeley College of Engineering (Executive Education)
• U.S. Army veteran, senior NCO in a Civil Affairs company

I am open to research collaboration on LLM moral evaluation, refusal behavior, and pluralistic alignment.

Contact: open an issue on any of my repos, or reach me through GitHub.