Building governance infrastructure for AI systems that make employment decisions — at the intersection of I-O psychology, employment law, and ML security.
Automated screening is effectively universal in large-employer hiring: 99% of Fortune 500 companies use an applicant tracking system, and 88% of employers acknowledge that their system screens out qualified candidates.1 Independent verification has not kept pace. When Cornell researchers measured compliance with NYC Local Law 144 — the first law anywhere requiring bias audits of automated employment decision tools — they found that of 391 employers surveyed, 18 had posted an audit report and 11 posted both an audit report and a conforming transparency notice.2
I build the methodology and tooling to close that gap.
I am the author of the AVS Framework (Audit • Validation • Security), a unified governance methodology for AI employment decision systems:
- Audit — adverse impact analysis under the Uniform Guidelines (29 C.F.R. 1607) and state bias-audit laws (NYC Local Law 144, Illinois HB 3773, Colorado SB 26-189)
- Validation — I-O psychology validity evidence (SIOP Principles, job analysis, KSAO mapping, differential item functioning) proving AI tools measure what they claim to
- Security — ML threat assessment aligned to NIST AI RMF 1.0 / AI 600-1 and the OWASP ML Top 10 (data poisoning, model extraction, prompt injection, drift)
- Governance — policy, human-oversight, and monitoring maturity assessment for sustained compliance
Most AI hiring audits check one of these four boxes. AVS checks all four in a single, repeatable, standards-mapped assessment.
How it is licensed. The Audit and Security pillars are released as MIT-licensed open-source software (below) — free to run, modify, redistribute, and deploy commercially, with no registration, key, quota, or restriction. The Validation and Governance pillars require job analysis and organizational assessment performed with a client, and are delivered as professional services through Nauta Research Labs, the research and consulting venture I founded.
avs-framework — open-source (MIT licensed) statistical engine implementing the Audit and Security pillars.
pip install avs-frameworkEvery release is archived to Zenodo with a citable DOI: 10.5281/zenodo.21806797
| Adverse impact | Four-fifths rule per 29 C.F.R. 1607.4D, pooled two-proportion Z-test, and automatic fallback to Fisher's exact test when expected cell counts fall below 5 |
| Severity model | Four-level classification that separates practical from statistical significance, surfacing the disparities that single-criterion audits miss |
| Name-swap bias testing | Black-box test for name-based bias in resume scoring, using fully synthetic resumes |
| Drift detection | Chi-square composition drift, Kolmogorov-Smirnov score drift, and per-group selection-rate drift between periods |
Runs on de-identified applicant flow exports from any ATS (Workday, Greenhouse, iCIMS, Lever) — no vendor source code or model access required. Tested against Python 3.10–3.13, with an 80% line-coverage floor enforced in CI.
Education
- M.S., Industrial-Organizational Psychology — Missouri State University
- Graduate Certificate, Statistics and Research Design — Missouri State University
Certifications
- SHRM-CP — Society for Human Resource Management
- MLSecOps — Protect AI
- Prosci ADKAR — change management
Peer-reviewed publications
- The Role of AI in Reducing Implicit Bias in Recruitment — IJARCST, 2024
- Evaluating an AI-Driven CAT Platform — IJIRCCE, 2025
- Green AI for Sustainable Employee Attrition Prediction — IJARCST, 2025
Citation record: Google Scholar
Technical reports — self-published, not peer-reviewed
- Dual-Endeavor Assurance for AI Employment Decision Tools — Nauta Research Labs, 2026
Peer review service
- Reviewer, International Journal of Research and Applied Innovations — large language model safety, red-teaming methodology, AI governance
Statistics & Psychometrics — adverse impact modeling, DIF / measurement invariance, CFA, SEM, validity coefficient correction, Cochran-Mantel-Haenszel
Engineering — Python (pandas, numpy, scipy), R, SPSS, Power BI; data pipelines for HRIS/ATS exports
Standards — Title VII, ADEA, ADA, Uniform Guidelines (29 C.F.R. 1607), NIST AI RMF, OWASP ML Top 10, SIOP Principles, AERA/APA/NCME Testing Standards
Footnotes
-
Fuller, J. B., Raman, M., et al. (2021). Hidden Workers: Untapped Talent. Harvard Business School Project on Managing the Future of Work and Accenture. ↩
-
Wright, L., et al. (2024). Null Compliance: NYC Local Law 144 and the Challenges of Algorithm Accountability. Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency. ↩