Skip to content
View j-arndt's full-sized avatar

Block or report j-arndt

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
j-arndt/README.md

Justin Arndt

Technical writer. Regulated systems, APIs, docs-as-code.

For eight years I wrote the controlled documentation for validated laboratory systems at a GSK vaccine manufacturing site: validation plans, requirements specifications, risk assessments, test protocols, SOPs, change records, and root cause investigations. More than 100 documents and thousands of pages, every one through formal review and approval, and the set held up through internal audits and FDA inspection preparation.

Then I learned Git and recognized the whole thing. Document control is docs-as-code with heavier tooling: one source of truth, reviewed changes, reasons attached to every revision, history that never lies. Now I write developer-facing documentation on the modern toolchain, with the discipline the regulated world beat into me.

Start here

j-arndt.github.io is my portfolio and my working proof. It runs like a production documentation project: Markdown in a public repo, every change through a pull request, and CI that lints prose with Vale against the Google developer style guide, checks structure and links, and fails the build on any broken reference.

Three entry points, depending on what you are hiring for:

You need Read
API documentation FileMaker Data API reference, built on a hand-authored OpenAPI 3.1 specification
Tutorials and onboarding Qdrant in Docker: a first vector search
Conceptual and architecture writing Anatomy of an audit trail

The short version of the pivot is an essay: Change control was docs-as-code before Git.

Background

  • Eight years as a technical writer and computer system validation engineer, onsite at a GSK vaccine manufacturing site through Alphanumeric Systems
  • Ten years supporting clinical imaging and PACS/DICOM systems at Lancaster General Health (Penn Medicine), using Epic daily from its 2007 go-live
  • M.S. in Data Analytics and B.S. in Cybersecurity and Information Assurance, Western Governors University
  • ITIL Foundation; CompTIA A+, Network+, Security+, CySA+, and PenTest+; Linux Essentials; Claris certifications at Expert level (App Developer for Claris FileMaker Pro, App Developer for Claris Connect, Claris FileMaker Server Administrator)

The security and data credentials are not decoration. They mean I can document authentication flows, access models, audit requirements, and data pipelines without waiting for an engineer to translate.

Contact

Reach me at [email protected] or on LinkedIn. Lancaster, Pennsylvania, working remote.

Pinned Loading

  1. agent-long-term-memory agent-long-term-memory Public

    Local LLM memory that survives VRAM exhaustion. Compresses sessions to 4KB using Bloom filters + Merkle trees, stores to Cloudflare R2, hydrates in <500ms. QLoRA bakes memory overnight.

    Python

  2. agent-resurrection agent-resurrection Public

    SEED_POLY encoding protocol: deterministic agent state compression using Bloom filters, Merkle trees, and DEPA resurrection probes. Drift score to 3 decimal places.

    Python

  3. certified-agentic-firewall certified-agentic-firewall Public

    Production-grade input/output firewall for agentic AI. Cryptographic audit trail on every tool call and model output. 21 CFR Part 11 and SOC 2 aligned.

    Python

  4. certified-phi-egress-firewall certified-phi-egress-firewall Public

    HIPAA-aware AI egress firewall. Detects 18 PHI categories and blocks outbound LLM traffic containing patient data. Sub-50ms latency, full audit log.

    Python

  5. ironclad ironclad Public

    Formal verification sandbox for the IRONCLAD solver featuring Lean 4 algebraic proofs, deterministic verification environments, and cryptographic execution audit logs.

    Lean

  6. miniVectorDB miniVectorDB Public

    An SRAM-optimized, mathematically exact, billion-scale MIPS vector database utilizing Fast Walsh-Hadamard Transforms and FP8 quantization bounds.

    Python