ML Engineer, Retrieval and Grounded Generation

About the role

Build retrieval and grounded generation for an AI-assisted platform in a secure government cloud, ensuring generated text is tied to cited source records.

What you will do

• Design prompts and output schemas and own model packaging, versioning, serving and rollback
• Instrument telemetry for retrieval and generation quality, overrides, abstentions, latency and throughput
• Provide bounded model assistance for narrative extraction with every output tied to its source passage
• Maintain a modular in-boundary serving path alongside the primary managed inference service

What they look for

• Five or more years of experience including a retrieval-augmented system you built yourself
• Strong Python with hands-on embeddings and vector retrieval at scale
• Ability to explain retrieval design, grounding tests, citation failures and rollback in detail
• Clarity on what actually shipped in past work
• US citizenship and an active US Secret clearance to start

Nice to have

• Deploying models into restricted or air-gapped environments
• Operating self-hosted or open-weight models
• Fine-tuning, adapters or custom embeddings
• Federal DevSecOps, RMF, ATO or DoW cloud experience