Assistant Research Scientist (PREP0004961

PREP Research Associate

Johns Hopkins, founded in 1876, is America's first research university and home to nine world-class academic divisions working together as one university.

$50,000-$70,000 a year

This position is part of the National Institute of Standards and Technology (NIST) Professional Research Experience Program (PREP). NIST recognizes that its research staff may want to collaborate with researchers at academic institutions on specific projects of mutual interest and, therefore, requires those institutions to be recipients of a PREP award. The PREP program involves staff from a wide range of backgrounds conducting scientific research across various fields.

Individuals in this position will perform technical work supporting the collaboration's scientific research.

Measuring Accuracy Uplift From LLM Adversarial Pressure/AI Research

Overview: ITL’s AI Program role in the project “Measuring Accuracy Uplift From LLM Adversarial Pressure/AI Research” involves the following tasks: (1) developing and evaluating an agentic deepresearch pipeline in which large language models (LLMs) autonomously research a corpus of authoritative documents and produce a fully cited report, and (2) designing adversarial LM-judge evaluation probes that measure whether the pipeline’s claims and citations are factually grounded in the source material.

Our success depends upon the availability of highly skilled domain experts. We are challenged with difficult tasks that require not only expertise in agentic AI systems and large language models, but also in designing evaluation methodologies for applications where measuring factual grounding under adversarial pressure has not been attempted before.

U.S. citizenship is preferred.

• Developing and tuning LM-judge evaluation probes that measure the factual grounding and citation quality of agentic AI outputs.
• Design adversarial pressure scenarios that stress-test an agentic research pipeline’s factual accuracy against a trusted document corpus.
• Analyze the accuracy uplift attributable to adversarial verification and compare that approach against baseline agentic pipeline performance.
• Produce high-quality publications based on research and results present at internal and external meetings and conferences.

• US citizenship is preferred.
• An M.S. or PhD degree in Computer Science with 3 or more years of relevant experience.
• Expertise in Python and state of the art AI models, including large language models and agentic AI systems.
• Ability to build deployable complex software solutions for agentic AI evaluation pipelines.
• Strong oral and written communication skills and strong presentation skills.

The referenced salary range represents the minimum and maximum salaries for this position and is based on Johns Hopkins University’s good faith belief at the time of posting. Not all candidates will be eligible for the upper end of the salary range. The actual compensation offered to the selected candidate may vary and will ultimately depend on multiple factors, which may include the successful candidate's geographic location, skills, work experience, internal equity, market conditions, education/training and other factors, as reasonably determined by the University.

Johns Hopkins offers a total rewards package that supports our employees' health, life, career and retirement. More information can be found here: (Use the "Apply for this Job" box below)..

The Johns Hopkins University is committed to equal opportunity for its faculty, staff, and students. To that end, the university does not discriminate on the basis of sex, gender, marital status, pregnancy, race, color, ethnicity, national origin, age, disability, religion, sexual orientation, gender identity or expression, veteran status or other legally protected characteristics. The University is committed to providing qualified individuals access to all academic and employment programs, benefits and activities on the basis of demonstrated ability,…