
Domain practitioners who capture real work, write acceptance conditions, and adjudicate agent runs—so evaluation records contain expert judgment, not chat labels.
Fellows are the judgment layer of Raycaster's evidence system—not a chat-labeling gig. Capture work, specify acceptance conditions, adjudicate runs.
How contribution works →You are the judgment layer of Raycaster’s evidence system: turn consequential workflows into tasks, rubrics, and verdicts others can inspect.
Capture representative work with de-identified source packets and edge cases
Specify rubrics and observable acceptance conditions a peer would defend
Review trajectories and artifacts the way a dossier or domain reviewer would
Adjudicate graders so public evidence and private evals stay honest
Use Raycaster for your own work, learn how AI is being applied to regulatory workflows, and connect with a small group of experienced peers.
Full Raycaster Individual Plan access (valued at $200/month)
Direct access to the product team
Early access to new workflows and features
Optional paid projects
Peer community + in-person meetups + dinner series
Invitation to Raycaster's Global Summit (2026)
~2–4 hours/month. Mostly async. Your judgment becomes part of an evaluation record.
Provide de-identified templates, source packs, and edge cases.
Help define rubrics and what a competent peer would accept.
Review trajectories, artifacts, and grader evidence with structured feedback.
No. This program only works with de-identified examples. Placeholders are fine.
If you've worked on real submissions and want to help shape AI for regulatory work, we'd like to hear from you.