Jon Norris

I work in AI evaluation, science education, environmental science, media QA, and field observation. I help test whether AI systems are accurate, useful, and reliable when they are used in real workflows.Through JNO Systems LLC, I review AI tools used in education and other knowledge-heavy settings. I look for the kinds of problems that are easy to miss at first glance, including hallucinated facts, weak source grounding, poor citation behavior, rubric mismatch, and outputs that sound confident but are not fully supported.My AI evaluation work has included RLHF, model review, agent reasoning, multimodal QA, and rubric-based scoring. That work has made me especially interested in how AI systems behave under ambiguity, partial information, and production pressure.Before moving deeper into AI evaluation, I spent more than 20 years in science education, environmental science, marine biology, aquaculture, ecological observation, and media workflows. I taught Biology, Environmental Science, Marine Biology, and Physics, and much of my work has always centered on evidence, systems, and real-world problem solving.I also bring production experience from Disney+ media delivery workflows, where small issues in video, audio, subtitles, metadata, localization, or IMF packages could create downstream problems. That background shaped how I think about QA: the final output is only as reliable as the process behind it.My field work now includes long-term hummingbird behavior research in the San Fernando Valley. I use video, photography, field notes, and repeat observations to study feeder behavior, territoriality, habitat use, and urban wildlife patterns.The common thread across all of this is reliability. Whether I am reviewing an AI response, an education tool, a media package, or a field observation record, I care about whether the evidence supports the output and whether the system can be trusted in practice.I am especially interested in AI model evaluation, science AI audits, EdTech review, environmental science data review, biodiversity and field-observation workflows, multimodal QA, hallucination and citation integrity, RLHF, human data workflows, and rubric design.