AI Software Engineer

AI Software Engineer

Location: Farmington Hills MI

Must Have

• 3-4 years of Document Control experience
• Author
• Taxonomy Development
• Technical Manuals
• Vocabulary

Job Description

We are looking for a Senior Technical Writer to join our AI agentic engineering team. You will develop and manage the knowledge catalog that our production AI agents read from authoring the entries that encode how our IT enterprise actually works, and structuring them so agents retrieve the right knowledge at the right moment.

This is a writing role at its core, but the reader is different. Your audience is an AI agent operating under a context budget, and behind it, the engineer who has to trust what that agent produces. Success is measured less by page count than by whether agents behave correctly because your entries were clear, correctly scoped, and easy to find.

What You'll Do

• Author knowledge entries — service overviews, runbooks, decision records, schemas, specifications, and glossary terms — drawn from workflows, policies, engineering standards, and technical documentation
• Write the short descriptions and summaries that drive progressive disclosure — the text an agent reads to decide whether loading the full entry is worth the context it costs
• Structure orientation paths so an agent onboarding to an unfamiliar domain encounters the mental model, entry points, and invariants first, and drills into detail only on demand
• Maintain hierarchy coherence across a growing catalog — cross-linking, indexing, reachability, and the layered structure that keeps entries discoverable as the corpus scales
• Own the scoping metadata that routes entries to the right consumer in the right context — precision here determines whether an agent gets the relevant standard or the wrong one
• Author and maintain prompt assets — reusable skills, agent instructions, and reference material consumed directly by production agents
• Codify engineering constitutions — turn review standards, security requirements, and platform conventions into structured, versioned rules that automated code review agents apply on every pull request
• Interview developers and engineers to capture the rationale behind decisions — the "why" is rarely present in the source material, and it is usually the part that matters most
• Advance entries through review and approval with engineering and governance partners, so only verified content reaches production consumers
• Govern vocabulary — maintain the glossary and naming conventions, resolving terminology collisions before they propagate into schemas, tooling, and agent behavior
• Keep the catalog current — identify and retire stale entries as the systems they describe change

Core Capabilities

• Writes with precision under hard length constraints, where an unnecessary sentence carries a measurable cost
• Structures information for machine consumption as fluently as for human readers
• Reads primary source material — code, configuration, infrastructure definitions, runbooks — accurately enough to summarize it without an engineer rewriting the result
• Interviews technical staff effectively and recognizes when an answer is incomplete
• Sustains consistency of voice, terminology, and structure across a large, cross-linked corpus
• Exercises editorial judgment about what belongs in the catalog, what should be merged, and what should be removed
• Partners credibly with developers, engineers, and governance teams without needing content pre-digested

What Differentiates This Role

Most technical writing roles optimize for a human who is scanning a page. This one optimizes for an agent deciding what to load, and a governance team deciding whether to trust it. A description is a retrieval decision. An index is a retrieval surface. A vague entry does not merely confuse a reader — it produces a wrong answer in a production system.

You will also write the prompt-side assets, not just the reference material, which means the boundary between "documentation" and "how the agent thinks" is genuinely yours to manage. Writers who thrive here think in information architecture first, care about how knowledge is consumed rather than only how it is published, and are comfortable being accountable for downstream agent behavior.