Senior AI Systems Engineer — Retrieval & Context

You know that the model is rarely the problem. The context is. You've built retrieval or context-management systems that real users depended on, you've been burned by a chunking decision or a context-ordering choice, and you measured your way out instead of guessing. You can explain why a hybrid search beat pure vector search on a particular corpus without reaching for a framework name. You care what the documents actually look like, and who's allowed to see them. BoodleBox's AI foundation — context assembly, retrieval, document intelligence, model routing — is shared by every experience we ship, across higher education, legal, and defense. You'd be the deepest person in the building on how information gets retrieved, selected, assembled, and handed to a model, and your job is to make that layer measurably better: quality, latency, cost, with the numbers to prove it. Whether you come from the context-orchestration side or the search-and-retrieval side, we want you if you understand the other well enough to work across the whole system. This is a senior individual-contributor role. You'll be the technical authority on retrieval and context. You won't be leading the AI Systems group — that's a separate path, and someone is already growing into it. If being the expert in the room matters more to you than the title, keep reading. This role is fully remote, with a company hub in Colorado Springs, CO.

What You'll Own

• Context assembly. What goes in the window, in what order, compressed how, when it won't all fit. Every product we ship depends on you getting this right.
• Retrieval architecture. Chunking, indexing, hybrid search, re-ranking, citation and grounding — in production, at scale, on messy real documents from universities and law firms.
• The evaluation harness. Repeatable datasets and metrics for relevance, groundedness, and answer quality, including whether a user ever sees a document they shouldn't. Decisions get made on these numbers.
• The quality-cost-latency trade. Including an honest read on where long-context models, agentic retrieval, or memory approaches actually help and where they're hype.
• Reusability. Retrieval and context capabilities that Chat, Classroom, Boxes, and the next vertical all use, rather than three versions of the same thing.
• Your services in production. You steward a set of AI services on Kubernetes with a named backup: they're healthy, documented, and safe for other engineers to change.
• The legal document workflows. Document processing and document creation on the shared foundation — the thing blocking a revenue line right now.

‍

Who You Are

• You've run retrieval or context systems in production. Real users, real scale, real regressions. Not a notebook, not a demo.
• You know how models actually behave. What context ordering does. What length does. What happens at the edges of the window.
• You measure. You can name your eval dataset, your metric, and the time the eval disagreed with your gut — and what you did about it.
• You've thought about permissions. Tenant boundaries and document access are part of retrieval correctness to you, not someone else's layer.
• You write production Python. Tested, packaged, observable, and you've been paged for it at least once.
• You can read Scala. Or you're willing to learn, because part of our platform lives there.
• You're motivated by depth. You'd rather be the person every engineer asks about retrieval than the person who runs the meeting.

‍

We're Not Checking Boxes. We're Looking for the Real Thing.

Didn't come up through a traditional path? Good. Tell us about it.

No big-lab logo on your resume? We've hired people who outperformed everyone who did.

Career gap? Weird pivot? Started somewhere completely left field? Bring it. The people who got here the strange way usually have the best instincts.

What we actually want to see: someone who notices when the answer was wrong because the context was wrong without thinking about it. If that's you, the path you took to get here is just a good story.

‍

Who Thrives Here

We move quickly, and the speed comes from engineers who own outcomes end to end. You'll ship retrieval improvements and see them in a professor's workflow the same week. Nobody will hand you a spec; you'll work with product architects who understand users and engineers who understand the platform, and the retrieval decisions will be yours.

You'll fit if you're curious about how students and faculty actually use AI, if you find messy document corpora interesting rather than annoying, and if you believe the quality of an AI answer is an engineering problem with an engineering solution. This is education. Getting it right matters more than getting it clever.

‍

What We Offer

• Competitive base salary of $175,000–$185,000, plus bonus and equity.
• Unlimited PTO.
• Medical, dental, and vision coverage. BoodleBox covers 75% of premiums.
• 401(k).
• Fully remote work environment.
• Real career growth. Inside a high-growth, mission-driven company.