Mercor Research Fellowship — APEX

ABOUT MERCOR

Mercor's mission is to organize human intelligence to power the AI economy. We're a leading AI data company, building the layer between human expertise and frontier models. Millions of domain experts on the platform are paid over $4 million per day to train frontier AI models. Mercor's APEX benchmark family measures AI's real-world impact on professional work. Mercor Enterprise brings this same infrastructure to Fortune 500 companies: helping companies capture how their best people actually work, translating that expertise directly back into agents.

Mercor is creating a new category of work where expertise powers AI advancement. Achieving this requires an ambitious, fast-paced and deeply committed team. You’ll work alongside researchers, operators, and AI companies at the forefront of shaping the systems that are redefining society. Mercor is a profitable Series C company valued at $10 billion. We work in-person five days a week in our San Francisco, NYC, or London offices.

About the Fellowship

Mercor’s APEX benchmark family measures whether frontier AI models can actually do economically valuable work: multi-hour agentic tasks in investment banking and corporate law, real professional accounting workflows, real-world software engineering, and graduate-level science. Every APEX benchmark is built and validated with Mercor’s network of domain experts — not written from a textbook.

The Mercor Research Fellowship funds people to build the next generation of benchmarks and evaluation techniques. You pitch a benchmark or eval methodology you want to build — a new domain, a harder task format, a better way to measure agentic reliability — and if selected, you get the time, compute, expert labor, and mentorship to design, implement, and release it end to end.

You’ll work directly with the APEX research team, get access to real enterprise evaluation problems from Mercor’s Fortune 500 and frontier-lab partners, and see your benchmark shape how the industry measures AI capability.

Program Details

• Duration: 3–6 months, rolling admission

• Commitment: minimum 30 hours/week; full-time preferred

• Location: remote, or in-person at Mercor’s San Francisco office

• Admission: apply with a specific benchmark or eval technique you want to build — the fellowship is funded around your pitch, not a generic research rotation

What You’ll Do

• Propose and scope a new benchmark or evaluation technique in a domain APEX doesn’t yet cover well, or a meaningfully harder version of one it does.

• Design task specifications and grading rubrics in partnership with Mercor’s network of vetted domain experts — lawyers, accountants, engineers, scientists, and consultants.

• Build and validate the benchmark: pilot tasks, calibrate scoring, and stress-test for contamination and gameable shortcuts.

• Run frontier models against your benchmark and analyze where and why they fail.

• Publish your results — as a paper, an open dataset, a new leaderboard on APEX, or a methodology the APEX team adopts internally.

• Partner with Mercor’s research and engineering teams to fold what you learn back into APEX’s public benchmark family.

Focus Areas

• Long-horizon, multi-app agentic tasks in professional services (law, finance, consulting) — extending APEX-Agents

• Real-world software engineering evaluation beyond issue resolution — extending APEX-SWE

• Professional accounting and finance workflows — extending APEX-Accounting

• AI-for-Science evals: research-level mathematics, biology, materials science, and theoretical physics

• Novel evaluation methodology: contamination resistance, rubric design, human-vs-model grading agreement, cost-adjusted scoring

• Strong pitches outside this list are welcome — we fund the best ideas, not the closest fit to a template.

What We’re Looking For

• Genuine interest in evaluation as a research discipline — not just a stepping stone to a model-building role.

• Background in CS, ML,

Responsibilities

• Propose and scope a new benchmark or evaluation technique in a domain APEX doesn't yet cover well, or a meaningfully harder version of one it does.
• Design task specifications and grading rubrics in partnership with Mercor's network of vetted domain experts — lawyers, accountants, engineers, scientists, and consultants.
• Build and validate the benchmark: pilot tasks, calibrate scoring, and stress-test for contamination and gameable shortcuts.
• Run frontier models against your benchmark and analyze where and why they fail.
• Publish your results — as a paper, an open dataset, a new leaderboard on APEX, or a methodology the APEX team adopts internally.
• Partner with Mercor's research and engineering teams to fold what you learn back into APEX's public benchmark family.

Requirements

• Genuine interest in evaluation as a research discipline — not just a stepping stone to a model-building role.
• Background in CS, ML