RAG AI Engineer for Classified Government Infra

Amentum; Hanover (MD); Hybrid; USD 195,000 - 250,000; Full time

Benefits offered by this job

Health, dental, vision insurance

401(k) matching

Educational reimbursement

Parental leave

PTO and holidays

Employee stock purchase plan

Job summary

Amentum is seeking an AI Engineer for Hanover, MD to design, build, and deploy AI capabilities that operate in classified and disconnected environments. The role focuses on retrieval-augmented generation, open-weight LLM fine-tuning, and secure serving on Government infrastructure.

You will work closely with customers and internal teams to translate mission needs into practical AI solutions, ensure compliance with RMF and ATO, and help advance responsible AI adoption across programs.

Qualifications

• Experience delivering RAG applications used by real users.
• Fine-tuning/adapting open-weight LLMs with measurable benchmarks.
• Hosting and serving models on GPU infrastructure with proper sizing.

Responsibilities

• Design, build, and deploy AI systems in production software.
• Bridge data science and software engineering for ML projects.
• Support ATO processes and secure, disconnnected deployments.

Skills

RAG design

LLM fine-tuning

Python

LangChain

LLM orchestration

Docker & Kubernetes

Security & compliance

communication to leadership

document ingestion

GPU hosting

Education

Bachelor's degree in CS/Engineering

Equivalent combination of education and experience

Tools

LangChain

LlamaIndex

Semantic Kernel

Docker

Kubernetes

Job description

Amentum is seeking an AI Engineer for Hanover, MD to design, build, and deploy AI capabilities that operate in classified and disconnected environments. The role focuses on retrieval-augmented generation, open-weight LLM fine-tuning, and secure serving on Government infrastructure.

You will work closely with customers and internal teams to translate mission needs into practical AI solutions, ensure compliance with RMF and ATO, and help advance responsible AI adoption across programs.