Remote AI training and data labeling jobs

Build, tune and evaluate ML models in Python with MongoDB-backed data pipelines for a customer AI training project, and document every experiment. Remote contractor, $80–140/hour, 35 openings. scikit-learn, TensorFlow or PyTorch plus hands-on MongoDB is the core stack.

Technical experts audit the tasks used to train and evaluate AI systems, checking each is accurate, realistic, solvable, reproducible and properly tested. You are matched to one of eight specialties, from GPU kernels and Kubernetes to CVE security and SWE-Bench, and take an assessment in it. $60/hour, remote contractor.

Build, break and verify real machine learning engineering tasks: model components, reproducible training and inference workflows, memory and throughput optimisation, numerical debugging. Output-based pay, 100 openings, global remote, roughly 15 hours a week at $100–150/hour.

Run simulated contract negotiations and redlining exercises, review how an AI handles the same scenarios, and write the grading rubrics that measure it. Three years of in-house technology-transactions experience is the stated bar. Task-based pay at roughly 3.5 hours per task.

Assess adolescent eating disorder cases in writing, applying DSM-5, EDE-Q and SCOFF, judging medical stability against MEED, and saying whether FBT or CBT-E fits the clinical stage. Wants a licensed clinician with a recent adolescent caseload, and accepts paediatric medicine alongside psychiatry and psychology. Remote contractor work.

Refine CRM data models, arbitrate conflicting operational policies and document why you ruled the way you did, and design revenue workflows that hold up under vertical regulation. Four years with your hands actually in the system is the bar; the ad states outright that management-only exposure does not count. Remote contractor work.

Redline commercial agreements (MSAs, NDAs, DPAs) against established playbooks, decide the company position and the acceptable fallbacks, and write the reasoning behind every markup. Asks for a JD, active US bar admission and four years of transactional practice. Remote contractor work on a project described only as confidential.

Advise on ab initio configuration interaction calculations of polarizability, critique atomic-structure work on Yb-171 and Yb-174, and explain sigma versus pi transitions, largely through spoken AI interviews, which on this listing are the work rather than the screening. Worldwide, 5 to 10 hours a week, and the ad asks only for what you can legally share.

Run simulated contract negotiations and redlining exercises, judge how an AI handles contract scenarios, and build the grading frameworks that measure it. Three years in-house on technology transactions (MSAs, NDAs, DPAs) is the stated bar, and the ad spells out what task-based pay means: roughly 3.5 hours per task.

Review and annotate data sets on children's learning and behaviour, judge model output for developmental and cultural accuracy, and write the guidelines that shape how an AI handles the linguistic, cognitive and psychosocial sides of child development. A PhD in the field is required; research or academic background counts, with or without clinical experience.

Review case studies and simulated patient interactions for military and veteran populations, then write the expert feedback, prompts and scenarios that teach a model what good clinical practice looks like, including the cultural nuance this population expects. Asks for a completed psychiatry residency or a psychology PhD, plus direct patient care experience.

Assess adolescent substance use case material, stage severity against DSM-5 SUD and ASAM criteria, and recommend a level of care, separating developmentally normal experimentation from clinically significant patterns, and flagging withdrawal, overdose and safeguarding risks. Wants a licensed clinician with a recent adolescent caseload, not AI experience.

Read adolescent case material and write structured suicide and self-harm risk assessments, applying C-SSRS and formal risk formulation, separating ideation from intent, and recommending escalation pathways and safety planning. The output trains models meant to guide non-specialists. Wants a licensed clinician with a recent adolescent caseload, not AI experience.

Read physician and patient research surveys and say whether the clinical terminology, treatment pathways and response options match how the disease is actually treated. A flat $200/hour, three years in a therapeutic area, and an MD is explicitly not required. Fully remote, on your own schedule.

Design and review physician and patient questionnaires (screeners, question wording, response scales, branching logic, respondent burden) and say whether an instrument will actually produce usable data. A flat $140/hour, and therapeutic-area specialism is welcome but not required. Fully remote, on your own schedule.

Build reinforcement learning environments that test whether an AI model can fix bugs, add features, refactor code and tune performance in real repositories, then write the golden reference solution it is graded against. Public open-source contributions are the requirement, not AI experience. Remote contractor work, roughly 15 hours a week.

Stress-test autonomous coding agents: review the code they write, find the edge cases where they break down, and build the rubrics and harnesses used to score them. Task-based, fully remote, 10–40 hours a week at your own pace.

Design the coding benchmarks that decide whether a frontier AI model can actually reason, debug and ship production code, and build the data pipelines behind those evaluations. A three-month remote hourly contract, full-time availability preferred.

Build and maintain the backend services, APIs and data pipelines behind AI training, evaluation and deployment. A fully remote hourly contract at 20–40 hours a week, aimed at engineers with five or more years of production experience.

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.

Build reinforcement learning environments that test whether an AI agent can fix bugs, implement features, refactor and optimise code while discovering and reasoning over information from real MCP servers. You design the reproducible environment, the deterministic verification and the golden reference solution. Remote contractor work, roughly 15 hours a week, hours of your choosing.

Review and draft complex transactional documents (M&A, private equity and capital markets agreements) and give the structured commentary that teaches a model how a BigLaw associate reasons, including on data privacy and cross-border deals. Asks for a US JD, active bar admission and AmLaw 100 experience. Remote contractor work at the widest pay band on this board.