Mercor AI Research Fund Grants
Mercor · Remote · San Francisco · Mexico | Mexico City · New York City · London ·
COMPENSATION
€•• – ••k
Listed on the employer's job board (Ashby)
- $5M total grant fund
- Covers API & expert costs
- US or UK based applicants
- Open license publication
- Non-exclusive IP rights
Mercor is committing $5 million to fund AI research that addresses the gap between benchmark performance and real-world utility. This grant supports projects involving realistic evaluations, high-fidelity environments, and long-horizon task execution.
Funding covers researcher hours, API costs, expert time from Mercor's platform, and conference stipends. Grantees retain ownership of their work while granting Mercor a non-exclusive commercial license.
What you'll do
- Conduct research on real-world AI evaluations and reliability
- Build interactive environments capturing work complexity
- Evaluate model planning and adaptation in long workflows
- Improve agentic capabilities like tool use and memory
- Develop robust evaluation methodologies for utility
- Publish findings as papers, datasets, or open tools
What you'll need
- Affiliation with university, non-profit, or small startup
- Organization must not sell AI training data
- Applicants must be based in US or United Kingdom
- Proposal must address a specific, well-scoped question
- Commitment to publish under open commercial license
- Acceptance of non-exclusive license terms for Mercor
Nice to have
- Experience with agentic evaluation methods
- Background in RL environments
- Expertise in post-training techniques
What you get
- Funding for researcher time and API costs
- Stipends for event and conference attendance
- Access to Mercor expert network for annotation
- Use of internal evaluation infrastructure
- Introductions to frontier lab researchers
Hiring process
- Submit one to two page Expression of Interest
- Include team background and research accomplishments
- Detail proposed project and expected outputs
- Provide timelines and resource requirements
- Avoid including confidential or proprietary info