AI Engineer

Amentum•Hanover, MD

28d•Hybrid

About The Position

Amentum is hiring an AI Engineer in Hanover, Maryland, to design, build, and deploy AI capabilities within classified and disconnected customer environments. The role focuses on retrieval-augmented generation (RAG), adapting and fine-tuning open-weight language models, and securely serving models on Government infrastructure. The engineer also helps customers and internal teams understand AI capabilities and adopt them responsibly. The position integrates machine learning models into production software and bridges data science with software engineering. Work includes data preprocessing, model selection, API development, and scalable infrastructure. The engineer translates mission requirements into AI architecture, evaluates models and tools within customer security constraints, and leads technical work across engineering, security, and mission teams while balancing delivery with compliance and governance. The work schedule is five days per week, eight hours per day. The environment is described as a hybrid or modern office equipped with high-performance computing infrastructure and cloud collaboration tools. Work involves cross-functional collaboration and may be fast-paced and high-pressure. An active TS/SCI clearance with polygraph is required. A bachelor's degree in a technical field is the stated minimum, although an equivalent combination of education and relevant experience may be considered. The position requires at least five years of professional experience in software engineering, machine learning engineering, or data engineering, including at least two years designing, building, and deploying LLM or machine-learning applications used by operational or business users.

Requirements

Active TS/SCI clearance with polygraph.
Bachelor's degree in Computer Science, Engineering, Data Science, Mathematics, or a related technical field, or an equivalent combination of education and relevant experience.
At least 5 years of professional experience in software engineering, machine-learning engineering, or data engineering.
At least 2 years designing, building, and deploying LLM or machine-learning applications relied on by operational or business users.
Knowledge of LLM fundamentals, including transformer architecture, tokenization, embeddings, context windows, inference parameters, hallucination, and prompt injection.
Knowledge of RAG design patterns, including chunking strategies, embedding-model selection, vector and hybrid search, reranking, and retrieval evaluation.
Knowledge of model adaptation methods, including supervised fine-tuning, LoRA, QLoRA, and distillation, and when to use them instead of prompting or RAG alone.
Knowledge of open-weight models such as Llama, Mistral, Qwen, and Gemma, including licensing, provenance, and supply-chain considerations for Government use.
Knowledge of security and compliance frameworks for AI in classified environments, including RMF, ATO, ICD 503, the NIST AI Risk Management Framework, and CUI and classified-data handling requirements.
Demonstrated experience designing, building, and deploying at least one RAG application for real users, including document ingestion, chunking, embedding selection, vector-store configuration, and retrieval evaluation.
Demonstrated experience fine-tuning or adapting open-weight LLMs and measuring results against a defined benchmark.
Experience hosting and serving models directly on GPU infrastructure using vLLM, TGI, NVIDIA Triton, or similar, including GPU sizing and quantization tradeoffs.
Experience delivering AI capabilities in disconnected or classified environments, including transferring models, containers, and dependencies across air gaps and supporting systems through ATO.
Experience building access-controlled retrieval pipelines that enforce clearance, compartments, and need-to-know restrictions.
Experience building LLM evaluations and guardrails, including accuracy testing, hallucination measurement, red teaming, and prompt-injection defense.
Experience building document-ingestion pipelines for large unstructured collections, including parsing, OCR, metadata tagging, and preservation of classification markings.
Proficiency in Python and at least one LLM orchestration framework, such as LangChain, LlamaIndex, or Semantic Kernel, as well as Docker and Kubernetes.
Ability to explain AI concepts, tradeoffs, and limitations clearly to technical and nontechnical audiences, including Government leadership.
Ability to translate mission requirements into AI architecture and choose suitable approaches for each problem.
Ability to evaluate emerging models and tools quickly and recommend adoption within customer security constraints.
Ability to lead technical work across engineering, security, and mission teams.
Ability to balance delivery speed with customer security, compliance, and governance requirements.

Nice To Haves

Prior experience as an AI Engineer, Machine Learning Engineer, Software Engineer, Data Engineer, or Solution Architect with direct build responsibility.
Experience supporting Intelligence Community, DoD, or Federal customers.
Experience deploying AI workloads in classified cloud regions, including AWS Secret or Top Secret, Azure Government Secret or Top Secret, or Government-operated GPU clusters.
Experience building agentic workflows that call tools, APIs, or other systems, including Model Context Protocol (MCP) or similar integration standards.
Experience with MLOps or LLMOps practices, including model versioning, CI/CD for models, monitoring, and drift detection.
Experience preparing or contributing to ATO packages for systems that include AI or machine-learning components.
Experience with enterprise AI platforms such as Microsoft 365 Copilot, Azure AI Services, or Amazon Bedrock in Government cloud regions.
Experience coaching teams on effective AI use and building business cases that quantify productivity gains from AI adoption.

Responsibilities

Design, build, and deploy AI systems by integrating machine-learning models into production software.
Develop RAG applications, including document ingestion, chunking, embeddings, vector or hybrid search, reranking, and retrieval evaluation.
Fine-tune or adapt open-weight language models and measure performance against defined benchmarks.
Host and serve models on GPU infrastructure, accounting for GPU sizing and quantization tradeoffs.
Deliver and support AI capabilities in disconnected or classified environments, including moving models, containers, and dependencies across air gaps and supporting systems through ATO.
Build access-controlled retrieval pipelines that restrict content according to users' clearance, compartments, and need-to-know.
Create evaluation methods and guardrails for LLM output, including accuracy tests, hallucination measurement, red teaming, and prompt-injection defenses.
Build document ingestion pipelines for large unstructured collections, including parsing, OCR, metadata tagging, and preservation of classification markings.
Translate mission requirements into working AI architecture and select appropriate approaches such as prompting, RAG, fine-tuning, or agents.
Evaluate emerging models and tools and recommend adoption within customer security constraints.
Lead technical work across engineering, security, and mission teams.
Explain AI concepts, tradeoffs, and limitations to technical and nontechnical audiences, including Government leadership.
Help customers and internal teams understand AI capabilities and adopt the technology responsibly.