What the work involves

You write, solve, and adjudicate hard quantum-optics problems that are used to benchmark and train frontier AI models. A typical task begins with a physical setup — a cascaded OPA chain, an SU(1,1) interferometer with unbalanced gain, a two-mode squeezed state passed through a lossy channel — and asks for a fully derived result: covariance matrices propagated through Bogoliubov transformations, sideband photocurrent spectra under homodyne detection, degradation of squeezing when loss is modeled with fictitious beamsplitters. Work splits across three tracks:

  • Solver — author original problems with unambiguous statements and complete, step-by-step reference solutions.
  • Auditor — check another contributor's derivation for algebraic errors, sign conventions, wrong vacuum-noise normalization, or claims that don't survive the lossless limit.
  • Adjudicator — resolve disagreements between solver and auditor and issue the final ruling on the canonical answer.

Much of the value sits in the failure modes: a model output can look impeccable and still botch a hyperbolic-identity simplification, drop a phase, or silently assume a symmetric loss that the problem didn't grant. Your job is to catch that and say precisely why it is wrong.

What the platform screens for

micro1's screen is AI-led and conversational, and it follows up. Expect to be asked how you would actually set up a covariance-matrix calculation, why a particular squeezing figure degrades the way it does under 10% loss, or what changes when the second amplifier's gain is detuned from the first. Generic familiarity is easy to detect and gets filtered; the screen rewards people who can name conventions they use, describe calculations they have personally carried through, and identify where a plausible-looking derivation breaks. A PhD (or clearly equivalent research output) in AMO physics, quantum optics, or quantum information is the baseline. No AI or machine-learning background is expected or asked for.

Logistics

Fully remote, contractor engagement, asynchronous. There are no fixed shifts; you claim tasks from a queue and deliver against per-task deadlines. Most contributors run 10–20 hours per week, though volume tracks the customer's project phase and can be uneven — periods of heavy queue followed by quiet stretches are normal. Pay in the $80–160/hr range has been observed for this role, with placement depending on track (adjudication typically sits higher than solving) and demonstrated depth; it is not a guaranteed rate.