Frontier AI Research Scientist: Benchmark & Evaluation

Hidden Jobs; United States; Remote; USD 250,000 - 400,000; Full time

Benefits offered by this job

Equity

Collaboration with leading tech labs

Conference publication support

Job summary

Hidden Jobs in the United States seeks a research scientist to advance frontier AI evaluation, training, and improvement. You will tackle benchmarks where existing methods fall short and pioneer new directions across evaluation, synthetic data, and agentic science.

The role emphasizes hypothesis-driven experiments, publication-quality results, and collaboration with top researchers to push the boundaries of AI reliability and verifiable reasoning.

Qualifications

• PhD or equivalent research experience in ML/CS/technical field.
• Proven ability to formulate and execute original research.
• Strong understanding of modern large language models and frontier AI.
• Excellent experimental design and technical writing.

Responsibilities

• Identify gaps in benchmark literature and propose ambitious new research programs.
• Design and build novel benchmarks with rigorous task generation, calibration and validation.
• Develop methods for generating synthetic STEM training data and study impact on performance.
• Investigate hallucination, uncertainty, calibration, and evaluation methods.
• Research AI systems with long-horizon capabilities including literature search, coding, tool use, and reasoning.
• Publish findings at leading AI venues.

Skills

Original research

Experimental design

Quantitative analysis

Technical writing

LLMs

Education

PhD or equivalent research experience

Job description

Hidden Jobs in the United States seeks a research scientist to advance frontier AI evaluation, training, and improvement. You will tackle benchmarks where existing methods fall short and pioneer new directions across evaluation, synthetic data, and agentic science.

The role emphasizes hypothesis-driven experiments, publication-quality results, and collaboration with top researchers to push the boundaries of AI reliability and verifiable reasoning.