Justin Mullins

AI/ML research scientist and technical lead focused on agentic systems, LLM behavior and evaluation, AI assurance, adversarial AI, RF machine learning, and formal approaches to reasoning in AI. Current work spans multi-agent research assistants, LLM knowledge systems, IC-wide AI observability requirements, and mission-relevant evaluation of emerging AI capabilities.

Background combines applied ML research, production-oriented software development, formal logic and epistemology, and leadership across national security research teams.

Associate Research Scientist, AI/ML - University of Maryland, ARLIS - College Park, MD

(2026-01)

• Lead and contribute to AI/ML research efforts involving AI agents, LLM evaluation, AI assurance, observability, and responsible deployment of emerging AI technologies for high-consequence missions.
• Architect and lead development of Open Analyst, a multi-agent research assistant for analytic tradecraft that supports open-source research, artifact retrieval, external connectors, and interactive deliverable generation.
• Lead LLM-wiki, a curated knowledge system for LLM research supported by multi-agent librarian workflows and agent-served knowledge surfaces.
• Define technical requirements and needs gaps for enterprise AI observability and assurance across intelligence community use cases.
• Mentor junior scientists, software engineers, and researchers on AI methods, model evaluation, agentic systems, and the practical limits of emerging technologies.
• Translate technical feasibility, risk, and model behavior into language usable by policy, mission, and technical stakeholders.

Principal AI Engineer / Data Scientist - MasterPeace Solutions, Ltd. - Columbia, MD

(2025-10 - 2026-01)

• Served as PI for an automated modeling, testing, and evaluation architecture for multimodal AI systems, including computer vision, speech-to-text, OCR, and machine translation.
• Designed evaluation workflows for complex AI systems, emphasizing repeatability, model comparison, performance characterization, and transition to secure technical environments.
• Coordinated engineering implementation across multimodal model testing, data pipelines, and secure GPU-based evaluation workflows.

Senior AI Technical Advisor - Cornerstone Defense - Washington, DC

(2025-08 - 2025-11)

• Advised senior intelligence community stakeholders on LLM adoption, technical risk, model provenance, evaluation, governance, and responsible deployment.
• Authored standards and guidance for LLMOps, model provenance, and evaluation that translated technical constraints into policy-relevant artifacts.
• Briefed technical and nontechnical stakeholders on feasibility, hype, risks, and assurance needs for emerging AI technologies.

Research Scientist, Machine Learning - National Security Agency - College Park, MD

(2022-01 - 2025-08)

Built and led TorchSig, an open-source RF machine learning library. Optimized performance-critical DSP and ML paths. Adapted deep learning models to RF spectrogram tasks.

• Built and led TorchSig, an open-source RF machine learning library, including dataset generation, spectrogram tools, transforms, training workflows, and GNU Radio integration.
• Optimized performance-critical DSP and ML paths in Rust through Python FFI to support high-throughput signal processing workflows.
• Adapted DETR, XCiT, YOLO-style models, transformers, and deep learning pipelines to RF spectrogram classification, detection, and emitter localization tasks.
• Designed RF anomaly detection approaches using discrete VAE-style encoding, masked modeling, and pattern-of-life learning for complex spectrum environments.
• Mentored junior researchers across ML, RF signal processing, software engineering, and research execution; managed local GPU resources and briefed leadership on technical progress.

Research Fellow, AI/ML and NLP - Center for Cyber Social Dynamics, University of Kansas - Lawrence, KS

(2024-05)

• Conduct research on LLMs, Decision Transformers, NLP, and large-scale social media and news analytics for modeling normative and social dynamics.
• Explore methods for extracting patterns from online information environments and connecting computational models to policy-relevant interventions.

Post-Graduate Researcher - University of Manchester - Manchester, UK

(2016-09 - 2018-01)

• Conducted research in philosophy of mathematics, formal logic, and the history of analytic philosophy while developing computational and data-analytic methods for research workflows.

Adjunct Professor, Philosophy and Logic - Park University - Remote

(2013 - 2016)

• Designed and taught courses in symbolic logic, analytic reasoning, argumentation, and philosophy for online students.

Sergeant, SIGINT Analyst / Electronic Warfare Operator - United States Marine Corps - Camp Pendleton, CA

(2003 - 2008)

• Conducted tactical SIGINT analysis, operated RF intercept and electronic warfare systems, and produced mission-relevant reporting for operational leadership.

Ph.D. - Philosophy - University of Kansas (2022)

M.A. - Philosophy - University of Kansas (2015)

B.A. - Philosophy - Wichita State University (2012)