Remote AI training and data labeling jobs

Arabic-speaking PhD chemists and biologists write specialised science prompts in Arabic and grade AI answers for accuracy and dual-use safety. Part-time remote at $38–42/hour; Saudi Arabia or MENA preferred, not required. 13 hires this month, the most active listing in the series.

Umbrella listing for Mercor's energetic materials red-team panel: chemists and engineers or operators write benign, dual-use and adversarial prompts, judge frontier AI replies against a policy standard, and write reference answers. Remote contract at $65–75 per task; 16 hired this month.

Write or verify 10-option multiple-choice benchmark questions in applied maths (signal processing, actuarial science, optimization, climate modeling and more), with chain-of-thought solutions and references. For maths PhDs and doctoral candidates. Remote, 10+ hours a week, $61–77/hour.

Author executable scientific-computing problems in ecology, biochemistry and genetics for Sci Code, a new AI benchmark: source a paper, dataset or repo, write the prompt and grading criteria, and keep it only if frontier models mostly fail. PhD plus Python or R, Git and Docker. 6 weeks, 20+ hours a week, $70/hour.

Solve, audit or adjudicate research-level benchmark problems in quantum optics: cascaded optical parametric amplifiers, SU(1,1) interferometers, loss and two-mode squeezing, using Bogoliubov and covariance-matrix methods. Ten openings, contractor, $80–160/hour, fully remote.

Turn hands-on CRISPR, cloning and genotyping know-how into expert content and assessments that train AI on molecular and cell biology. Contractor, remote, 100 openings, $70–90/hour. A bachelor's degree plus real wet-lab gene-editing experience is the bar.

Build realistic biostatistics benchmark tasks (clinical trial analysis, regulatory review, observational studies) with real datasets and 35+ item grading rubrics. Paid per accepted task, quoted at $60–100/hour. MS or PhD plus four years in pharma, CRO, hospital or academic medicine. 50 openings.

Red-team frontier AI models on chemical misuse: write benign, dual-use and adversarial prompts from forensic casework, judge the model's answers against a policy standard, and write reference answers. Remote contract at $65–75 per task.

Solve, audit or adjudicate research-level problems in holography: AdS/CFT, on-shell gravitational actions, the Weyl-Fefferman-Graham gauge and tensor algebra of boundary curvature invariants for holographic Weyl anomalies in d ≤ 8. Ten openings, contractor, $80–160/hour, remote.

Annotate biological datasets, write problem sets and case studies grounded in real medicinal chemistry, and check AI answers for scientific accuracy. Contractor, remote, 30 openings, $90–120/hour. A PhD, PharmD or MSc and drug-discovery experience are preferred.

Solve advanced physics problems from your own specialism and write them up in LaTeX, verified with SymPy, Python and Jupyter, for AI training. Open worldwide across a dozen subfields, 10–15+ hours a week over an 8–10 week project. 14 openings, contractor, $100–200/hour.

Benchmark problems where error correction meets disordered stat mech: random-bond Ising and Ashkin-Teller models, the toric-code threshold, the Nishimori line, Kramers-Wannier duality and 4-state Potts numerics. Worldwide, part-time 5–10+ hours a week. Ten openings, contractor, $100–200/hour.

The UK posting of Mercor's molecular biology project: design primers, plasmids, gRNAs, mRNA constructs and repair templates as ground truth for a frontier model, and write the rubrics that judge them. PhD strongly preferred, first-author record expected, 20 hours a week. $70–105/hour.

Design the primers, plasmids, gRNAs, mRNA constructs and repair templates that become ground truth for a frontier model, and write the rubrics that judge sequence design quality. PhD strongly preferred, first-author record expected, 20 hours a week minimum. US only, $70–105/hour.

Solve advanced computational fluid dynamics problems (meshing, turbulence modelling, multiphase flow, heat transfer, solver setup), then judge whether AI-generated answers are engineering-sound and write up what is wrong when they are not. Three years of validated simulation work is the bar, not AI experience. Remote contractor engagement.