What the work involves
You build the evaluation material, not just consume it. A typical task starts with you inventing a technically realistic scenario from your track — a ring resonator FSR and Q budget, a metasurface phase-gradient design, a link budget with dispersion and nonlinearity limits, an antenna impedance-match problem — then authoring the expert reference solution and a grading rubric precise enough for someone else to apply consistently. Other tasks flip the direction: you receive a submitted solution and judge whether the physics holds, whether the numbers are plausible, and whether the approach reflects real domain judgment or a confident-sounding shortcut. Simulation work is frequently part of the loop, so tool fluency in something like Lumerical, Zemax, COMSOL, HFSS, Meep, or Tidy3D matters more than breadth of tool names.
What the platform screens for
- Hands-on modeling recency. Screeners probe for setups you actually ran: mesh and boundary conditions, convergence checks, where your model disagreed with measurement and what you did about it.
- Named tool depth. Expect follow-ups on the specific solver you claim — PML settings, symmetry planes, port definitions, ray vs. wave regimes.
- Problem-design instinct. Can you write a question that is hard for the right reason, has a defensible unique answer, and cannot be pattern-matched from a textbook?
- Grading calibration. Whether you can separate a wrong final number from a wrong method, and score partial credit without drifting.
Logistics
Fully remote and fully asynchronous — no standing meetings, no fixed hours. Work arrives as discrete tasks you accept and complete on your own schedule, which makes it workable alongside a staff engineering or research position. Throughput expectations are per-task rather than per-week, though contributors who take longer stretches away tend to see fewer routed tasks. Pay is stated by the platform at $100/hr with milestone bonuses for sustained quality; treat both as observed rather than guaranteed.