AI Engineer

## The Role

We're hiring an **AI Engineer** to design and ship secure, LLM-powered services for U.S. government missions. You'll build the retrieval, the guardrails, and the evaluation harness that let a government team actually rely on a model's output, not just be impressed by a demo. The work runs on Azure OpenAI inside environments with real constraints, and the people on the other end are doing serious work. Getting it right matters.

## What You'll Do

- Build RAG pipelines end to end: ingestion, chunking, retrieval, grounding, and the plumbing that keeps answers tied to a source a reviewer can check
- Design guardrails and safety layers so a service behaves inside its lane under adversarial and messy real-world input
- Stand up model evaluation: offline eval sets, regression checks, and the honest measurement that tells you whether a change helped or just felt better
- Deploy and operate LLM services on Azure OpenAI in accredited environments, with the logging and access controls those environments demand
- Work shoulder to shoulder with government stakeholders and our design team so the thing we ship fits the mission, not a generic benchmark

## What We're Looking For

- Hands-on experience building LLM applications that left the notebook: RAG, prompt engineering, tool use, and the evaluation to back it
- Strong software engineering in Python, with comfort reading and writing the surrounding system (APIs, data stores, pipelines)
- A real feel for where models fail and the instinct to measure it rather than hope
- U.S. citizenship and the ability to obtain a government clearance; residence in the Eastern or Central time zone

## Strong Advantages

Nice-to-haves, not strict requirements.

- Production time with Azure OpenAI or Azure AI services
- Experience shipping into accredited or regulated environments (FedRAMP, IL-level, or similar)
- Background in retrieval, search, or applied ML beyond prompting an API

## Why This Role Is Special

Most AI roles ask you to make something sound smart. This one asks you to make it trustworthy, for users who can't afford a confident wrong answer. You'll own the parts that decide whether a government team keeps using the tool after the novelty wears off, and you'll do it on a small team where your judgment shows up in the product.

## Location

Remote in the US, Eastern or Central time zone. U.S. citizenship required.

## Next Steps

Tell us about an AI feature you made genuinely reliable and how you proved it. One real example beats a polished resume here.