Founder
2026 – present- Designed and shipped Inqura, an evidence-analysis platform for investigations and compliance review, and ConKurrence, a multi-model evaluation harness. Both are described under Systems built.
Board-certified clinical psychologist who has spent three decades building and overseeing behavioral-health systems, from military screening and suicide surveillance to national healthcare inspection, and who now builds and evaluates AI tools for evidence-based review.
I'm Joe Etherage. My work has three threads. As a clinician, I led behavioral-health departments and programs for the U.S. Air Force, the U.S. Army and NATO's Supreme Headquarters Allied Powers Europe. As a builder, I designed and ran measurement systems: post-deployment screening that assessed 41,000 soldiers, and a suicide-event report that the Department of Defense adopted department-wide. As an overseer, I have directed healthcare inspection work that tests whether systems do what they claim, most recently on generative AI in clinical care.
In 2026 I founded JE Vectors LLC and built Inqura, an evidence-analysis platform for investigations and compliance review, along with ConKurrence, a harness for measuring whether AI judgments hold up. I am interested in AI oversight, healthcare and mental-health AI governance, and evaluation, as an employee or as a consultant.
An evidence-analysis platform for investigations and compliance review, designed and built solo with Claude Code as the engineering agent. Every claim in a finding cites its evidence, every finding carries a confidence level, and the server blocks sign-off until the reviewer has opened the cited evidence.
I wrote the original ten-criteria evidence-evaluation framework, then validated it down to the one-page guide that runs in production: two judgments per piece of evidence (is it relevant; does it have a material limitation), with an explicit materiality threshold. In controlled multi-model tests, rater agreement rose from chance level to substantial. An independent reproduction confirmed the agreement result but not the expert-anchor check, and I am correcting that public claim.
A multi-model evaluation harness (command line, MCP server and web interface; 194+ automated tests) that measures whether AI judgments agree with each other and with expert anchors, so accuracy claims can be tested rather than asserted.
Program lead for post-deployment behavioral-health screening at Madigan and Fort Lewis: 41,000 assessments and 16,000 referrals, extended to Schofield Barracks, Hawaii, in 2009.
Helped develop the Army Suicide Event Report, the Army's suicide surveillance record. The Department of Defense built its department-wide DoD Suicide Event Report from it, launched January 2008.
Clinical champion and product owner of the Army web application the SELF program ran on. Presented as a model for the military electronic health record's behavioral-health module (2008).
Co-author of a leader-completed screening instrument, with manual, that fed soldiers into behavioral-health referral.
Reports of the U.S. Department of Veterans Affairs Office of Inspector General, Office of Healthcare Inspections. These are institutional publications; I contributed to each as a director in that office.
Meritorious Service Medal (one oak leaf cluster) · Commendation Medal · Achievement Medal · Global War on Terrorism Service Medal · Outstanding Unit Award (three oak leaf clusters)