AI Researcher

Fort George G. Meade Complex, MD
Pay Plan: GG , Grade: 07/0 to 15/10
Open: , Close:

Job Posting: 1263706

Responsibilities include:

• Design and run experiments that characterize AI system capability, behavior, and failure modes; build benchmarks and evaluation methodology where none exist.
• Develop training, fine-tuning, retrieval, and inference techniques suited to Agency data, scale, and security constraints.
• Prototype novel AI applications and agentic systems that put emerging capabilities in analysts' hands, and iterate them based on user feedback.
• Investigate model internals, robustness, and safety, including how AI systems can be attacked, manipulated, or induced to fail.
• Assess vendor, open-source, and internally developed AI systems against government baselines and mission requirements.
• Track the external state of the art and provide critical, evidence-based assessments of new models and techniques.
• Communicate results through technical reports, briefings, prototypes, and internal publications; mentor junior researchers and interns.

As an AI Researcher at NSA, you will investigate, characterize, and extend what AI systems can do for the mission. The work spans rigorous evaluation and method development (benchmarking, training approaches, model analysis) through applied prototyping that turns emerging capabilities into working tools. You may be assigned to a research organization, embedded with a mission element, or enrolled in a development program that rotates you through several.

In each case you will work alongside NSA experts in AI, mission subject matter, and related technical fields, with opportunities to participate in internal technical round tables and external conferences.

The qualifications listed are the minimum acceptable to be considered for the position.

Degrees must be in Computer Science, Software Engineering, Computer Engineering, Data Science, Information Systems, Information Technology, Mathematics, Applied Mathematics, Statistics, Applied Statistics, Operations Research, Engineering, Computational Linguistics, Quantitative Finance, or Physical or Biological Sciences; all degrees must include coursework in linear algebra, multivariable (vector) calculus, and programming.

Relevant experience must include one or more of the following: deep learning, neural networks, generative AI, or computer vision.

ENTRY/DEVELOPMENTAL Entry is with an Associate's degree plus 2 years of relevant experience, or a Bachelor's degree and no experience, or a Master's degree and no experience.

Additional experience may also include research and/or development in any technical area.

FULL PERFORMANCE Entry is with an Associate's degree plus 5 years of relevant experience, or a Bachelor's degree plus 3 years of relevant experience, or a Master's degree plus 1 year of relevant experience, or a Doctoral degree and no experience.

Additional experience may also include research and/or development in any technical area.

SENIOR Entry is with an Associate's degree plus 8 years of relevant experience, or a Bachelor's degree plus 6 years of relevant experience, or a Master's degree plus 4 years of relevant experience, or a Doctoral degree plus 2 years of relevant experience.

Experience must also include research and/or development in AI.EXPERT Entry is with an Associate's degree plus 11 years of relevant experience, or a Bachelor's degree plus 9 years of relevant experience, or a Master's degree plus 7 years of relevant experience, or a Doctoral degree plus 5 years of relevant experience.

Experience must also include research and/or development in AI and experience in leading people or projects.

AI/ML
- Deep learning frameworks (PyTorch, JAX, Tensor Flow)- Transformer architectures, large language models, and multi-modal systems
- Model training, fine-tuning, and parameter-efficient adaptation; preference optimization (RLHF/DPO)- Retrieval-augmented generation, agentic architectures, and tool use
- Inference and serving stacks (vLLM, Ollama, TGI); quantization (GGUF, AWQ, GPTQ)
Research & Evaluation
- Experimental design: controls, baselines, ablations, threat-to-validity analysis
- Applied statistics: significance testing, confidence intervals, power analysis, effect size
- Evaluation methodology: metric design, adversarial and held-out test-set construction, human evaluation protocols
- Model analysis: probing, attribution, interpretability, behavior and representation…