Teaching Frontier AI
to Reason with Rigor.
I help large language models reason more rigorously. As a subject matter expert engaged through micro1 on behalf of a frontier AI laboratory, I wrote the reference answers, prompts and grading guidance that a model's legal reasoning was measured against — and then went on using those tools in a working legal practice, which is where you find out what an evaluation missed. Twenty-five years of peer-reviewed research and a decade running a federally funded research program stand behind every rubric, dataset and review.
Dawn Marie Hayes, Ph.D. · Subject Matter Expert & Legal Research Consultant
A Rare Combination of Scholarly Rigor & Expert Judgment — Applied to AI
Dawn Marie Hayes, Ph.D. is a subject matter expert who works on both sides of applied AI: evaluating how well frontier models reason, and building AI-assisted workflows for professionals who depend on them. In the summer of 2026, engaged through micro1 on behalf of a frontier AI laboratory, she designed and completed more than 65 adversarial tasks built to challenge a model's legal reasoning, and authored the "golden answer" reference solutions, prompts and grading guidance behind them. She was not applying someone else's rubric; she was writing the standard the model was measured against.
What she brings to that work is twenty-five years of Ph.D.-level investigative and analytical practice, combined with hands-on legal research experience. Trained to spot inconsistencies, identify gaps, verify sources with precision, and synthesize large volumes of material into clear, actionable analysis — exactly the judgment model evaluation rewards. A quarter-century of university teaching also means fluency in assessment, rubric design and rigorous source verification, refined as a tenured Full Professor of History at one of New Jersey's largest comprehensive state universities.
Alongside this work, Dawn runs Quattro Sante, LLC, an independent legal research and drafting practice supporting New York and New Jersey attorneys on litigation, business, real estate and matrimonial matters — keeping her evaluation work grounded in how legal work is actually done, not how it appears in a training set.
Domain-Expert AI Evaluation
Grading frontier model output — with deep specialization in the legal domain — for accuracy, citation integrity, and reasoning quality.
Rubric & Assessment Design
Twenty-five years of university-level assessment and rubric design, applied to scoring AI output against professional standards.
Ph.D.-Level Analytical Rigor
Three books and fourteen peer-reviewed articles and chapters; a career of multi-source investigation and evidence synthesis.
Grounded in Active Practice
An ongoing legal research practice keeps AI evaluation connected to how expert work actually happens.
Rigor You Can Check
Analytical judgment is easy to claim. These are the parts with public evidence behind them.
$399,754 in Federal Funding
Principal Investigator on two National Endowment for the Humanities awards (2019–2021 and 2024–2027), carrying full responsibility for budget management, hiring, contracts and federal reporting.
A 602-Site Research Database
Founder and director of The Norman Sicily Project: 602 documented sites, roughly forty fields each, with GeoNames and Wikidata identifiers, parallel English and Italian annotation, and a draft-and-review editorial workflow that records provenance. Used by about 2,100 people worldwide each month.
A Seven-Discipline Team
Directs software engineers in the US and Italy, an earth scientist, a mathematician, UX designers and roughly 75 student researchers since 2015 — convening and reporting to a ten-member international advisory board.
A Peer-Reviewed Record
Three books and fourteen peer-reviewed articles and chapters, including work in Speculum. Manuscript reviewer for Al-Masāq: Journal of the Medieval Mediterranean; external reviewer for departmental reviews and faculty promotion cases at other institutions.
How I Help Train Better AI
Domain-expert services for AI laboratories and technical teams building and evaluating models on complex, high-stakes reasoning tasks.
Model Evaluation & Grading
Review and validate labeled documents. Grade model answers on complex reasoning tasks — including litigation, contract, and legal research — for accuracy, citation integrity, and sound reasoning.
Hallucination & Error Detection
Identify the inaccuracies, gaps, and hallucinations that separate competent AI from unreliable AI, and author corrected reference answers.
Rubric, Dataset & Prompt Design
Develop prompts, rubrics, and high-quality datasets representing realistic professional workflows to train models toward expert-level performance.
Expert-in-the-Loop Review
Bring professional-grade subject-matter judgment to human-in-the-loop training, RLHF, and model-safety review pipelines.
Assessment Is the Job. AI Just Changed the Classroom.
Grading expert reasoning at scale isn't a new skill for Dawn — it's a twenty-five-year practice. Designing rubrics, evaluating high-stakes analytical writing and verifying sources under peer review translate directly into the judgment a frontier laboratory needs: knowing what a correct answer looks like, and being precise about exactly where a model's reasoning breaks down.
Paired with hands-on professional practice — currently in law, across litigation, real estate, business and matrimonial matters — that scholarly rigor stays grounded in how expert work is actually practiced, not just how it appears in a training set.
Start a Conversation- 🤖Legal Research Consultant — micro1
- ⚖️The Legal AI Fundamentals Certification
- ✦AI for Lawyers: Communication & Creativity
- ✦AI for Lawyers: Time & Tasks
- 📚Professional Paralegal Mastery of Lexis®
Education, Certification & Experience
Education
AI & Legal Certifications
Licenses & Appointment
Subject Matter Expert, Legal Reasoning — micro1, on behalf of a frontier AI laboratory · Remote · June–August 2026
Legal Research & Paralegal Support
Alongside AI evaluation work, Dawn runs Quattro Sante, LLC — an independent contract legal research and paralegal practice serving attorneys and law firms across New York and New Jersey, including AI workflow consulting on tools such as Anthropic's Claude, Microsoft Copilot and LexisNexis Protégé.
Litigation Support
Trial transcript analysis, evidentiary summaries, and factual chronologies from pleading through trial.
Real Estate
Transaction documentation and title-related research for NY & NJ residential and commercial matters.
Business
Entity documentation, contract review support, and legal research for firms and solo practitioners.
Matrimonial
Divorce-related research, evidentiary review, and case-support materials.
Recent engagement: Contract Legal Research Consultant — HeitmannLaw, Staten Island, NY (November 2025–June 2026), supporting litigation, business, real estate and matrimonial matters in New York and New Jersey.
Let's Work Together
Whether you're an AI laboratory or technical team looking for domain-expert evaluation and training, or an attorney who needs research or paralegal support — reach out. Dawn responds promptly to all inquiries.
Based in the New York City metropolitan area. AI evaluation work is fully remote; legal research consulting serves attorneys and law firms across New York and New Jersey.