One workflow at a time. Each step ends with something your team can use and a number you can check.
We pick a workflow your team repeats every week. I collect real examples, write down what a correct result looks like, and test AI against them.
You get a short report with the error rate, the cases it gets wrong, and a clear call on whether it's worth building.
I build the workflow with your team, on the tools you already have, or on a simpler setup that replaces a paid tool when that costs less. It runs on real work while we compare time and accuracy against the bar from step one.
You get a working workflow, written instructions, and pilot results.
I train your staff to use it and to review what it produces, with a checklist, spot checks, and clear rules for what goes to a person. Then I check the results on a schedule, so quality doesn't drop.
You get trained staff, a review routine they own, and regular quality check-ins.
A good fit for teams that handle a lot of documents, email, forms, or reviews, and need the results to be right.
Describe one task your team repeats every week. I'll reply with whether AI can do it well and how I'd test it.
Roles in AI program management, human data operations, training and quality, and AI data delivery.
Or email owen.onderdonk@gmail.com for my resume.
Program management for human data and AI evaluation, built on hands-on review and evaluation work.
When reviewers of AI training data disagree, what can an automated check settle, and what has to stay with a person?
A framework I designed and published. It checks each claim in a review against the written instructions and the evidence the reviewer had, and returns one of four results. Unclear instructions go to the person who owns them and don't count against the contributor.
A system design, a synthetic demonstration, and a research base of sourced claims. Nothing here estimates accuracy on real work.
Read AutoQASince 2024
Entered frontier-model data work in February 2024 creating complex mathematics and physics training data. Advanced to contributor review and reviewer oversight, and in February 2025 expanded into project development, client collaboration, program management, quality leadership, and broader operations. Project details are confidential.
15 years
Built an engineering and advanced-manufacturing program spanning multiple schools. Coordinated the School Board, superintendent, administrators, and community stakeholders to define, support, and launch it. Increased STEM participation by 33% and helped students earn 200+ professional certifications. Trained 100+ professionals to adopt programming and computer-science tools as a Code.org facilitator.
NC State and Randolph College
Bachelor's in Physics (NC State), with experimental nuclear-physics research involving ultracold neutrons. Master's in Curriculum and Instruction (Randolph College).
✦ Learning Voyage LLC. Independent practice.