The work

A frontier lab is assembling a technical-expert panel for a model training-and-evaluation engagement, and the shortlist has to be vetted before anyone is onboarded. You take candidates who have already cleared a paper screen and spend about twenty minutes each confirming that the depth is real: that the person who lists CUDA kernel optimization can talk about occupancy and memory coalescing without reciting a blog post, that the vulnerability researcher can describe a bug they actually found and how they triaged it, that the compiler engineer knows what an IR pass they wrote was for. You are not running a hiring loop or a coding test — you are calibrating a claim against a rubric and writing the summary the program team acts on.

Each interview ends in a tier: advance, hold, or pass, plus a short evidence-based write-up. Alongside the technical read, you confirm the logistics that quietly sink engagements — availability, work authorization, exclusivity commitments — and flag anything that looks off, including credential inflation, rehearsed answers, or a candidate whose stated hours don't survive a follow-up question. Volume comes in bursts tied to the shortlist, and scheduling coordination is part of the job.

What the screen looks for

Mercor's screen is AI-led and will push on specifics. Expect it to ask how many interviews you've run, in what function, and at what seniority, then follow up on a single case in detail — what the candidate claimed, what you asked, what you concluded, whether you were right. It is testing whether you can hold a warm, on-time conversation with a senior engineer who may know more than you about their subfield, and still form a defensible judgment. Rubric discipline matters more than domain mastery here: you don't need to write GPU kernels, but you do need to know the difference between a candidate who is vague and one who is simply describing something you don't know.

Logistics

  • Fully remote, roughly 10 hours per week, weekday business hours for scheduling overlap.
  • Observed band for this listing is $50–60/hr; rates on Mercor vary by engagement and are not guaranteed.
  • Contract engagement tied to a specific panel build, so duration follows the shortlist.
  • Expect to work inside the client's rubric and summary format rather than your own.