# AI Technical-Doc Proofreader (Claude Opus)

> Catch typos and brand-rule violations across 500-page technical PDFs.

**Status:** Proof of concept  
**Client:** An enterprise manufacturer (details confidential)

## Problem

Proofreading 500-page technical manuals by hand is slow and misses image-rendered defects.

## What it does

A Claude Code proofreader for large technical publications. It walks a PDF page by page applying brand and term rules, hyphenation, and callout-vs-label checks, using Opus vision to catch glyph-path and image-rendered defects the text layer misses, and auto-chunks PDFs over 32 MB. No API keys and no Python pipeline to maintain: tech authors run it from Claude Code and get a reviewer-ready findings table. A documented model comparison locks the tool to Opus after Sonnet returned zero findings where Opus caught real defects.

## Measured results

- **551 pages** proofread in one pass — 47 MB of a dense technical publication, full coverage.
- **Opus > Sonnet** validated model lock — Opus caught brand-consistency defects where Sonnet returned zero on real pages.

## Stack

Claude Opus (vision + text), Claude Code, pypdf (chunking), Markdown findings report

## Topics

AI, LLM, Agents

---

Part of the proof-of-work portfolio of Michael Wahl — VP, Software Engineering & AI Enablement.

- Canonical project page: https://connect.michaelwahl.org/project/pdf-proofreader
- Full professional profile: https://connect.michaelwahl.org/
- Machine-readable context: https://connect.michaelwahl.org/llms.txt