Case study / 2025 / Architect and engineer
Automated market-report generation
A multi-agent pipeline that rebuilds PDF market reports as schema-marked, indexable HTML at scale.
Problem
A large library of market reports existed only as PDFs. They were invisible to search engines, could not be reused as web content, and represented locked-away value.
Constraints
Output had to be accurate (no fabricated figures), consistently styled (a fixed component library, not AI-invented layouts), and search-friendly (valid structured data, crawlable data tables behind every chart).
System
A multi-stage agent pipeline ingests each PDF, enhances content, generates HTML against a hand-built component library, and runs a QA pass that validates figures and structure before publish. Component selection is deterministic, so layouts never drift, and every chart is backed by a crawlable data table with Schema.org markup.
Outcome
241 reports were rebuilt as schema-marked, indexable HTML pages, turning locked-away PDFs into reusable content that a search engine can actually read.
Stack
A multi-agent generation pipeline, a hand-built HTML and CSS component library for deterministic layouts, and Schema.org JSON-LD with crawlable data tables for every visualization.
Table B - Outcome
- reports rebuilt as indexable HTML 241
- agent pipeline with a QA gate 5-stage
- charts backed by crawlable data tables 100%
Stack: Multi-agent pipeline / Schema.org JSON-LD / HTML/CSS component library