Senior Technical Writer & DocOps Engineer | Healthcare IT, AI Agent Tooling & Deterministic Documentation Systems
Portfolio & 14 Enterprise Samples: justinsevern.com
LinkedIn: linkedin.com/in/justinsevern
Location: St. Petersburg, FL (Open to Remote Staff/Senior DocOps, Technical Writing & AI Knowledge Roles)
Critical enterprise documentation rots faster than human teams can maintain it by hand—and in regulated healthcare software (HIPAA, Medicaid, HL7 FHIR, ANSI X12), stale docs or leaked PHI in screenshots cause real operational and compliance failures.
I build deterministic documentation tooling paired with bounded AI agent workflows (Claude Code skills + Model Context Protocol servers) to govern a 1,900+ article healthcare knowledge base:
graph LR
A[Jira Sprint / Git PR] -->|JQL & Schema Check| B(doc-type-classifier)
B -->|Microseconds / No LLM| C{13 Doc Archetypes}
C -->|Assigns Rubric| D[doc-skills-library]
D -->|Claude Code + 3 MCP Servers| E[Audit, Score & Variable Lint]
F[Staging UI] -->|Playwright Fail-Closed PHI Mask| G(smartclient-capture-kit)
G --> E
E -->|Credential & GA Scan| H[Document360 API Publish]
| Repository | Stack | Purpose |
|---|---|---|
| doc-skills-library | Markdown / Claude Code / MCP | 21 public, sanitized Claude Code agent skills (from 61 internal production skills) for auditing a 1,900-article corpus, scoring drafts, and linting system variables. |
| smartclient-capture-kit | JavaScript / Playwright | Automated UI screenshot capture for enterprise SmartClient apps with fail-closed DOM/canvas redaction so patient PHI or customer PII never ships in an image. |
| doc-type-classifier | Python (Zero-Dependency) | Classifies documentation into 13 structural archetypes from raw HTML in microseconds—no LLM calls, no network latency, 100% explainable decisions. |
- Release Notes Automation: Cut biweekly Jira-to-HTML release notes effort from 1 full workday (8 hrs) to 2 hours across 10+ consecutive on-time releases.
- Corpus Throughput: Scaled annual knowledge base maintenance from 88 edits/year to ~1,000 edits/year, authoring 779 new articles in 2026.
- Support Deflection: Classified 17,500 support tickets to map avoidable customer contacts to documentation gaps, saving 875+ support hours/year.