What the work involves

You build the test material that AI systems are measured against. In practice that means constructing realistic electrical engineering scenarios — a power distribution redesign under a load-growth forecast, an EMC failure discovered late in certification, a PCB respin driven by a signal integrity margin, a control loop that destabilizes at temperature extremes — and then writing the reference output a competent senior engineer would produce. Each task ships with a rubric: explicit criteria that separate defensible engineering judgment from confident-sounding textbook recall.

Coverage spans power systems and grid infrastructure, semiconductor and PCB design, embedded systems and firmware, signal processing, and control systems. Scenarios often reference the tools real teams work in — Altium, Cadence, MATLAB/Simulink, SPICE variants, and enterprise PLM/EDA stacks — and the governance around them: UL, FCC, and IEC certification paths, design review gates, DFMEA and reliability analysis, cross-functional sign-off.

What the platform screens for

  • Verifiable depth. Mercor's screen is AI-led and follow-up heavy. Vague answers about "working on power systems" get probed until they resolve into a specific design, a specific constraint, and a specific decision you made.
  • Ownership, not adjacency. The bar is direct responsibility for designs, certification programs, or systems at scale — typically 5+ years at a large technology, industrial, or energy manufacturer.
  • Evaluation judgment. Can you articulate why an answer is wrong, not just that it is? Rubric writing rewards people who can name the failure mode in someone else's reasoning.
  • Standards literacy. Practical knowledge of how compliance and design governance actually run inside a large organization, including where they slow things down.

Logistics

Fully remote and asynchronous. Contributors typically commit 10–20 hours per week, self-scheduled, with output measured in completed and accepted tasks rather than logged hours. Engagements run in project waves — throughput can be uneven, and calibration rounds early on tend to be slower per task while you learn the rubric conventions. Compensation observed in the $70–80/hr range; final rates are set by Mercor per contributor and per project and are not guaranteed.