AI Systems Engineer

For the OPS Consulting team, ‘the power to help’ means helping our clients, helping serve the mission, helping our employees and their families, and helping the community. Headquartered in Hanover, MD. OPS Consulting has over two decades of experience specializing in the most mission-critical operations. We are thought leaders and innovators. The ingenuity of our developers, engineers, cyber experts, linguists, and analysts is dedicated to empowering our clients, fulfilling The Mission, and remaining trusted leaders and advisers in national security and technology solutions.

We are looking for an AI Systems Engineer to join in Annapolis Junction, MD.

• Perform requirements decomposition for AI-enabled systems, translating mission‑level key performance parameters into hierarchical requirement structures (system requirements, sub‑requirements, design specifications) with threshold/objective tolerance bounds at each level, traceable to stakeholder needs
• Develop and maintain requirements traceability matrices and dependency maps linking concept of operations, system requirements documents, interface control documents, and test plans to ensure complete bidirectional traceability across AI system artifacts, including configuration management of AI system baselines, model versions, and governance specifications
• Design and implement multi‑agent orchestration patterns applying the separation of concerns principle, isolating planning, execution, and governance into independent architectural components where the executing element cannot self‑authorise, and compliance is verified independently
• Build and maintain structured input pipelines using retrieval‑augmented generation, hybrid search architectures, and context ranking algorithms to optimise AI model performance against mission‑specific requirements, including data pipeline assessment, training data quality evaluation, and input provenance verification
• Implement confidence‑based escalation logic with defined thresholds for autonomous operation, human‑in‑the‑loop intervention, and mandatory halt conditions, applying the KPP threshold/objective model to agent decision boundaries
• Design and integrate mandatory termination mechanisms (software, hardware, deadman, and scope‑boundary) for all autonomous agent deployments, ensuring human override capability at all operational levels
• Support AI TEVV activities, including developmental test planning for non‑deterministic systems, adversarial testing, model behavioural boundary verification, prompt injection defence assessment, and readiness scoring using structured green/yellow/red evaluation criteria
• Develop and maintain runtime monitoring components that enforce complete mediation of agent actions, detect behavioural anomalies, and generate cryptographically signed evidence chains with forward integrity
• Perform structured gap analysis and technical debt quantification for AI programmes using document cross‑referencing, entry/exit criteria validation, and compliance delta assessment against SE standards and programme baselines
• Contribute to the centralisation of programme terminology through controlled vocabulary development, ensuring consistent definition and usage of technical terms across AI system documentation and cross‑functional teams
• Support rapid prototyping efforts within gated development cycles, including intake scoring, feasibility assessment, milestone gate reviews, and validation framework execution across technical, user, security, and transition readiness dimensions
• Prepare and deliver technical documentation including concept of operations, system architecture descriptions, interface control documents, and test plans for AI‑enabled systems aligned to defence acquisition milestone requirements (SRR, PDR, CDR)
• Assist in the development of AI practitioner training materials covering structured prompting techniques, retrieval‑augmented workflows, and multi‑agent automation design for defence and intelligence practitioners
• Support ML model lifecycle management including performance monitoring, acceptance criteria validation, retraining trigger identification, and model versioning within…