What the work actually involves

You are the delivery owner for an LLM training pod: a team of 20 or more Python, JavaScript, Java (and long-tail language) developers writing, reviewing, and refining data used for supervised fine-tuning and RLHF. Day to day this means staffing and sequencing work, running review layers that catch bad prompts and buggy reference solutions before they reach the client, diagnosing why a batch's rejection rate spiked, and reporting status to researcher stakeholders who care about quality and turnaround more than process narrative. Projects arrive unstable — ambiguous rubrics, shifting task specs, uneven annotator calibration — and your job is to move them to a steady state and keep them there.

This is not a purely managerial seat. Turing explicitly wants someone who will open the repo, read the submissions, and find the quality problem themselves — unclear question phrasing, methodology drift, unreproducible results, or plain broken code. Expect to spend real hours in code review and spot-audits alongside forecasting, escalation, and client calls.

What the screen looks for

The evaluation is roughly 120 minutes across two rounds: one technical, one leadership. The technical round probes whether your hands-on ability is current — code quality judgment in at least one of Python, Java, or JavaScript, and the ability to explain what makes a training example good or useless. The leadership round probes scale and accountability: how many engineers you have actually managed, how you handled a missed delivery, how you calibrate reviewers who disagree, and how you talk to a dissatisfied client. Specific numbers, named languages, and concrete failure stories carry this screen; abstract management philosophy does not.

Logistics

  • Contractor assignment, 10-month duration, no medical or paid leave.
  • Three commitment tiers: 20, 30, or 40 hours per week; minimum 4 hours per day.
  • At least 4 hours of daily overlap with PST is required regardless of tier.
  • Open to candidates located in India and LATAM countries.
  • Pay band undisclosed by the platform; ask for the rate and tier mapping before the leadership round.