ML Engineer, Retrieval & Grounded Generation
About this role
Defcon AI needs an engineer to solve complex retrieval and grounded generation challenges within an AI-assisted platform operating in a secure government cloud. The analytics and AI engineering team builds a system that processes external records, matches them to the correct individuals, highlights priority items for human reviewers, and explains recommendations in defensible terms.
In this role, the candidate will construct embeddings, manage vector storage at scale, integrate language models, and enforce strict citation bounds to prevent unsupported text generation. Candidates who thrive here possess deep technical clarity regarding production retrieval-augmented generation systems, understand infrastructure management from model packaging to telemetry, and hold the required security clearances for government cloud deployment.
What you will do
- Implement embeddings, vector storage, and retrieval across a large, provenance-tracked evidence base.
- Integrate language models so generated text is bound to cited source records and rigorously test for citation failures.
- Design prompts and output schemas to support structured model interactions.
- Manage model packaging, versioning, serving, and rollback procedures for production infrastructure.
- Instrument telemetry for retrieval and generation quality, recommendation attribution, overrides, abstentions, and grounding failures.
- Provide bounded model assistance for difficult narrative extraction where deterministic processing falls short.
- Maintain a modular in-boundary serving path, self-hosted or managed, to ensure platform availability.
- Supply recorded rule context to every model-assisted step so outputs carry exact rule versions.
What you need
- Bring at least 5 years of professional experience, including a production or near-production retrieval-augmented generation system built firsthand.
- Demonstrate the ability to speak in detail about retrieval design, vector store selection, grounding tests, citation failures, and rollback mechanics.
- Demonstrate strong Python programming capabilities with hands-on experience in embeddings and vector retrieval at scale.
- Provide clarity on what actually shipped in past work, distinguishing between prototypes, proposals, and deployed code.
- Hold US citizenship and possess an active US Secret clearance to start.
Nice to have
- Experience deploying models into restricted or air-gapped environments
- Self-hosted or open-weight model operation
- Fine-tuning, adapters, or custom embeddings
- Federal DevSecOps, RMF, ATO, or DoW cloud environment experience
- Active Top Secret clearance
Benefits
- Competitive salary, bonus, and equity package
- 100% employer paid, comprehensive health insurance including medical, dental, and vision for you and your family
- Unlimited PTO, with your manager's approval
- Flexible work environment where you manage your work day
- 14 weeks of fully-paid parental leave
Skills and tech stack
Questions candidates ask
What is the remote policy for this role?
This is a fully remote role, with occasional travel to DEFCON AI HQ, customer sites, and vendor facilities as required.
What is the salary range for this position?
The typical salary range for this position is $165,000 to $200,000 based on experience, skills, and other factors.
What security clearance is required to start?
US citizenship is required along with an active US Secret clearance to start.
What experience is required with retrieval-augmented generation?
Applicants need 5+ years of experience, including a production or near-production retrieval-augmented system built yourself, along with the ability to speak in detail to your retrieval design.