This is a first-of-its-kind AI research project focused on helping AI systems better understand and complete real-world tasks in the social services and public benefits domain. Mercor is recruiting US-based experts with hands-on experience navigating these systems — social workers, benefits advisors, case managers, patient advocates, and anyone who regularly helps people access Medicaid, SNAP, SSDI, disability applications, low-cost medical and mental health care, or health insurance plans.

Your job is not to answer AI questions — it is to design and execute tasks that reflect how people actually interact with AI when seeking public benefits help. You will claim tasks across key social services categories, write realistic prompts for those tasks, execute the task yourself and provide evidence of completion, then write a detailed evaluation rubric to assess how well AI performs on that task. The goal is to close the gap between what AI can do in theory and what it can actually do in practice when a real person needs real help. A minimum commitment of 10 hours per week is expected. You will need a desktop or laptop computer (Chromebooks are not supported). This is an ongoing project with long-term potential and a rate of $40–$100/hr depending on task complexity and expertise.

Mercor is recruiting for a first-of-its-kind AI research project focused on everyday real-world tasks. AI is already strong at answering factual questions — but it struggles with the practical, daily-life activities that occupational therapists, life coaches, SEND educators, support workers, personal organisers, rehab specialists, and community professionals help people with every day. Your expertise will help close that gap.

You will claim tasks across 100+ real-world archetypes — daily planning, routines, running errands, community activities, meal preparation, household management, and more — write a clear and realistic prompt for each task, execute the task yourself while recording your screen, then write a detailed evaluation rubric to assess AI performance. A minimum commitment of 15 hours per week is required. You will need a desktop or laptop computer (Chromebooks are not supported). This is an ongoing project with long-term potential at $50–$200/hr depending on task complexity.

Mercor is collaborating with a leading AI lab to contract detail-oriented generalists for a large-scale data annotation project. As a Generalist Expert, you will support the development of AI systems by categorising and labelling diverse datasets using predefined taxonomies. The scale of this project is significant — 2,159 people were hired into this role in the most recent month alone — so barriers to entry are relatively low compared to specialist roles, but the consistency and quality of your annotation work matters.

Your core responsibilities will include synthesising information from large volumes of data, annotating and categorising text, images, and other data according to detailed guidelines, applying predefined rubrics and taxonomies to produce structured outputs, flagging inconsistencies and ambiguities in datasets, and contributing to AI model improvement through consistent work. The role requires a college degree and strong critical reasoning, reading comprehension, and written communication skills. Prior experience applying rubrics, taxonomies, or standardised guidelines is preferred but not required. Expected commitment is approximately 20 hours per week. Payments are weekly via Stripe or Wise.

Mercor is seeking detail-oriented Search Generalist Experts to support a high-impact project with a leading AI research lab. In this role you will help evaluate and improve how advanced AI systems perform on real-world search and browsing tasks. The core work involves assessing AI-generated search responses for factual accuracy, helpfulness, clarity, completeness, and overall quality — as well as evaluating whether the AI used search tools appropriately and formed effective queries.

You will compare model responses side by side and provide concise, defensible rationales. You may also write and refine prompts, golden answers, rubric criteria, and edge cases for search-related evaluations. Applying project guidelines consistently across ambiguous, multi-step, and real-world search tasks is central to the role, as is identifying recurring failure modes and escalating unclear cases to project leads. This is a strong fit for people with excellent written English, strong online research skills, and comfort making nuanced quality judgements at scale. Experience in search quality, fact-checking, trust and safety, content evaluation, annotation, or LLM evaluation workflows is preferred but not required. Multilingual skills are a plus. A bachelor’s degree is preferred. Payments are weekly via Stripe or Wise.