Write realistic feature and bug-fix tasks for AI coding models, plus deterministic verifiers that accept valid solutions and reject the rest. About 15 flexible hours a week, paid per accepted task on a $30–100/hour band, 100 openings; a strong testing background is the stated plus.
Write and review Lean 4 proofs for a leading AI lab, formalize informal mathematics and judge whether a model's proof actually proves the right statement. A W-2 part-time employment position through Cincinnatus LLC, at least 20 hours a week and up to 40, paying $90–110/hour.
Embedded firmware, FPGA/RTL, flight software and hardware test automation engineers review and write hard problems about software that runs on real hardware (timing, race conditions, interfaces, bring-up) for an AI research initiative. 5+ years, 3 recent hands-on. Remote contract, $100–120/hour.
Maintainers of real open source repositories review LLM-generated patches against their own codebase, justify each verdict in writing, and peer-review other experts. Open globally, about 15 hours a week, paid per accepted task at a stated $150–300/hour equivalent. 10 openings.
Software engineers rebuild technical PowerPoint slides, document system architectures and engineering processes, and review technical storytelling, so AI learns to explain engineering clearly. US, UK or Canada. 7 openings, contractor, $150–350/hour.
Build benchmark tasks for Terminus Bench 4 from real terminal and infrastructure problems: a problem setup, a working environment, a reference solution and an automated verifier for each. Remote independent contractor work through Meridial at $35/hour.
Experienced IT and systems managers: critique IT strategy, architecture, vendor and security-compliance scenarios and write the expert input that trains AI tools for IT professionals. 5+ years leading technical teams or enterprise infrastructure. 5 openings, contractor, $60–75/hour.
Work through real Stripe workflows (customers, payments, invoices, subscriptions, refunds, disputes) while recording your screen, then write prompts and rubrics from what you did. Needs current Stripe access; prior human data work preferred. Forty openings, $90–170/hour.
Remote hourly contract for Android developers who already run Android Studio on their own Mac. $55–65/hour, paid weekly via Stripe or Wise. The software is free; you need a Mac with a display above 2.5 megapixels (MacBook Pro 14" and 16" screens qualify). Tasks are not described.
Remote hourly contract for Mac power users who run macOS Sonoma 14.5 as their everyday OS. $60–70/hour, paid weekly via Stripe or Wise. The version is specific, so anyone who has upgraded past Sonoma may not qualify. MacBook Pro 14" and 16" screens meet the display requirement.
Audit AWS serverless and infrastructure-as-code tasks for a frontier AI lab: multi-service designs (Lambda, API Gateway, DynamoDB, EventBridge, Step Functions), CDK or CloudFormation fidelity, IAM boundaries and retry semantics. For US engineers with 3+ years building on AWS. $70–90/hour.
Despite the title, a CAD and computational-geometry role: turn dimensioned engineering drawings into validated 3D solids built in Python (build123d, CadQuery and similar) and exported as STEP, for an AI training project. Open globally, about 15 hours a week, per-task pay. 10 openings, $50–100/hour.
Competitive programmers write original problems absent from existing online judges, with a reference solution, a robust test suite and C++ checkers for multi-answer outputs, as RL environments for AI. 15+ hours a week, paid per accepted task against a $45–65/hour band. 300 openings.
Fix bugs, build features and refactor real codebases in Python, Java, Rust, C++, Go or TypeScript, then write coding tasks and feedback that test AI models. Competitive programming and open-source experience wanted. Paid per accepted task, $50–100/hour band, 100 openings, remote contractor.
A short, well-paid sprint for very senior software engineers: help a leading foundation-model lab improve its models on hard SWE tasks. 10+ years at top US tech firms, about 20 hours a week for 2–3 weeks. $150–210/hour by geography and level; the 2-hour vetting exercise is paid $100.
Work through real Google Analytics (ideally GA4) tasks on screen recording, then write prompts and rubrics from them so AI agents learn the workflows. Needs your own current GA access and prior human-data work. 40 openings, contractor, $30–100/hour.
Design GPU programming tasks in CUDA, WebGPU or GLSL for training LLMs on performance and architecture, including kernel profiling and C++ host code. Remote contractor, $60–95/hour paid per accepted task, 25 openings. Graphics, HPC and ML acceleration backgrounds all qualify.
US-based full-stack engineers with 3+ years build working software on a leading AI lab's pre-release models, integrating APIs, tool interfaces and evaluation harnesses and reporting where the models fail. Two language ecosystems required. Full-time W-2 through Cincinnatus LLC, 40 hours a week, $50–65/hour.
Two to three paid one-hour video interviews with Mercor about how enterprise security work is done and judged, feeding a benchmark for AI cyber defense agents. $125–175/hour, US-based, no prep and nothing to label or submit.
Remote hourly contract for people who run Ubuntu 24.04 as their everyday desktop OS: Linux users, sysadmins, DevOps and open-source developers. $60–70/hour, paid weekly via Stripe or Wise. No licence cost, but you need a display above 2.5 megapixels. Tasks are not described.
A hybrid, Bay Area-based W-2 role embedded with a leading AI lab: senior software engineers vet model outputs, write instruction specs and golden solutions, and build engineering benchmarks. 4+ years, senior-level progression and a CS or engineering degree. 40 hours a week for an initial 6 months, $65–105/hour.
Create, solve and review hard full-stack engineering tasks, debug across browser, API and database layers, judge AI-generated solutions, and port whole builds from one language to another. Global and fully remote, about 15 flexible hours a week, paid per task on a $50–100/hour band, 300 openings.
A paid expert conversation for engineers who build and run LLM agents in production: a short AI screening interview (no coding), then, if selected, a 30-minute live call on agent reliability, evaluation and internal adoption, paid $100–500 depending on depth of experience.
Sales ops and CRM people: carry out real HubSpot workflows (contacts, deals, pipelines, automation) on screen recording, then write the prompts and rubrics AI agents are trained and judged on. Needs your own HubSpot access. 40 openings, contractor, $28–92/hour.
US-based systems analysts bring requirements gathering, BPMN or UML process modelling and SQL troubleshooting to the design and validation of AI training scenarios. Contractor, remote (US only), 10 openings, $60–120/hour. Five years as a systems analyst or technical consultant.
Security-minded engineers debug vulnerabilities, audit codebases, run penetration tests and refactor code in Python, Java, Rust, Go, C++ or TypeScript to produce AI training data. About 15 hours a week, flexible, paid per accepted task against a $30–100/hour band. 300 openings.
Expert C# and .NET developers with strong MySQL skills build and maintain backend services and APIs on a micro1 project framed as AI training. Remote contract, only 2 openings, $60–110/hour.
Backend engineers, ideally with an AppSec background, write realistic coding tasks for AI (features and bug fixes) plus deterministic verifiers that accept valid solutions and reject the rest. About 15 hours a week, flexible schedule, paid per accepted task against a $30–100/hour band. 300 openings.
Build reinforcement learning environments where an AI agent must fix bugs, add features or optimise code using real MCP servers, with deterministic verifiers and golden solutions. About 15 hours a week, your own schedule, paid per accepted task against a $60–120/hour band. 100 openings.
Contribute real-world code, explanations and code reviews as AI training data, drawing on five-plus years in Python, Java, JavaScript/Node.js or full-stack work. A narrow $60–75/hour band, 15 openings, remote contract with documentation and architecture writing at its core.
Restaurant staff and managers who know Toast inside out review POS workflows, write scenario guides, and annotate real menus, orders and transactions so AI understands restaurant operations. Forty openings, contractor, a wide $28–92/hour band.
Design reinforcement learning environments where an AI agent must use real Model Context Protocol (MCP) servers to fix bugs, build features, refactor or optimise code, with deterministic verification and golden reference solutions. About 15 hours a week, paid per accepted task on an $80–120/hour band, 100 openings.
Create, solve and review hard software-engineering tasks in real repositories, including porting whole builds between languages and reviewing AI-generated patches. Global remote contractor, $100–150/hour paid per accepted task, about 15 hours a week. A verifiable record of substantial open-source contributions is required.
GPU specialists profile and optimise CUDA kernels, refactor C++ and CUDA code, and write GLSL and WebGPU shaders on a project run with a leading AI lab. Remote contractor role, 50 openings, $60–100/hour.
Tier 2/3 IT support people turn their troubleshooting into AI training data: diagnose hardware, software and network faults across Windows, macOS and Linux, document the fix in ticket form, and spot recurring patterns. Three years of Tier 2/3 experience preferred. $30–55/hour, 50 openings, contractor, remote.
Power users of Word and PDF (tracked changes, version control, forms, data extraction, digital signatures) contribute document expertise to AI training. A Master's, PhD or JD plus three years of intensive professional use is required. US-based preferred, 49 openings, $20–55/hour.
Build reinforcement learning environments that test whether AI models can fix bugs, add features, refactor and optimise real code, each with a reproducible setup and a golden reference solution. Around 15 hours a week, paid per accepted task on a $100–150/hour band, 100 openings, no AI experience needed.
Package real bug fixes, features, refactors and performance problems as reproducible reinforcement learning environments with golden reference solutions for AI coding models. About 15 hours a week, paid per accepted task on a $50–100/hour band, 100 openings. Same text as a higher-paying sibling posting.
Audit repository-level software engineering benchmark tasks for a frontier AI lab: reference patches, test harnesses, Docker isolation, and signs of answer leakage or reward hacking. For US engineers with 3+ years and real open-source contributor or maintainer history. $70–90/hour.
Remote hourly contract for Mac-based IT admins, sysadmins, DevOps engineers and cross-platform developers who already run VMware Fusion on their own Mac. $55–65/hour, paid weekly via Stripe or Wise. Requires a display above 2.5 megapixels; tasks are not described.
Remote hourly contract for power users, IT administrators and desktop support staff who run Windows 11 Pro as their daily OS. $60–70/hour, paid weekly via Stripe or Wise. Home edition users may not qualify, and a single 1080p screen does not meet the display rule. Tasks are not described.
Run 20-minute structured interviews that vet senior engineers (GPU kernels, security research, ML compilers) for a frontier lab's AI training and evaluation panel, then tier each candidate and write a summary. For experienced recruiters and technical vetters. US, part-time, about 10 hours a week, $50–60/hour.
QA engineers, SDETs and test automation engineers review browser-based test workflows for AI-generated web apps, checking that each test is feasible, isolated, deterministic and gives a clean pass or fail before it joins a benchmark dataset. 3+ years and Playwright, Cypress or Selenium experience. Remote hourly contract at $30–60/hour.
US-based cloud and DevOps engineers design and grade AI training tasks on Kubernetes failure diagnosis, AWS service integration, Terraform or CDK design and CI/CD, and write the rubrics behind them. 4+ years at a top-tier organisation. Full-time W-2 through Cincinnatus LLC at a leading AI lab, $75–110/hour.
Turn your favourite open repositories into reinforcement learning environments for frontier coding models: subtle bugs, non-trivial features or performance problems, each with a reproducible setup, robust verifiers and a reference solution. About 15 hours a week, paid per task on a $50–100/hour band.
Audit end-to-end AI-assisted coding sessions (traces from tools like Cursor, Copilot or Claude Code) used to train and evaluate a frontier lab's models, judging correctness, workflow and reasoning with rubric-based feedback. 3+ years of software development plus hands-on agentic coding. US-only, $70–90/hour.
ML systems engineers write and evaluate training tasks for a frontier lab across GPU kernels, performance profiling, distributed debugging and LLM inference serving, plus the rubrics that grade them. 2+ years of hands-on ML infrastructure work. Canada, UK or US; 40 hours a week. $90–120/hour.
Remote hourly contract for game developers, technical artists and virtual production or archviz specialists who already run Unreal Engine on their own Windows PC. $30–40/hour, paid weekly via Stripe or Wise. The engine is free to download; a display above 2.5 megapixels is required.
New roles land most days. Get them on Telegram or Discord as they are added, or read how the platforms pay before you apply.