What the work actually involves
You take on discrete 3D tasks — a hard-surface asset, a UV and texture pass, a lighting setup in Cycles or Eevee, a short animation, a procedural node network — and deliver both the .blend output and a clear written record of how you got there. The documentation is not an afterthought: scene setup, modifier stacks, node graphs, render settings, and the reasoning behind each choice are as much the deliverable as the model. Expect to troubleshoot too — fireflies, non-manifold geometry, UDIM mishaps, render times that balloon on a scene that should be cheap — and to explain the diagnosis, not just the fix.
Because this is AI evaluation work, the value sits in the trace. A model can produce plausible-looking Blender instructions; what it cannot reliably produce is the judgment about why a subdivision surface fails at that crease, or why a texel density mismatch shows up only at grazing angles. Your job is to make that judgment legible.
What the screen looks for
- Verifiable production history: named projects, pipelines you worked inside, formats you shipped in (glTF, FBX, USD, Alembic).
- Depth in at least one specialisation — hard surface, organic, texturing/UV, lighting and rendering, animation, or procedural/geometry nodes — held up under follow-up questions rather than a shallow tour of all six.
- Whether you can write down a workflow another artist could reproduce without you in the room.
- Realistic availability against a short, front-loaded engagement.
Logistics
Fully remote and asynchronous, no fixed hours. Projects are offered individually and you accept or decline; the expected commitment is 10–20 hours per week across a 2–3 week run. Pay has been observed at $40–50/hour, project-based, and is not guaranteed at any particular point in that band. A bachelor's in digital arts, animation, game design or similar is listed, with equivalent professional experience or a portfolio accepted in its place.