Case Study — Central Bank Intelligence (BOJ)

Problem

Central bank releases are dense, frequent, and spread across web pages + PDFs. Analysts need:

• fast summaries
• structured extraction
• evidence they can cite
• repeatable yearly reporting

What we built

A pipeline that:

• collects releases by year
• extracts text (including PDFs)
• runs multi-stage LLM extraction/summarization
• generates a yearly report + structured JSON outputs

Why it’s different

• Designed for long workflows (bounded concurrency, timeouts, retries)
• Built for review (run IDs, stage logs, provenance hooks)
• Optimized for throughput (async I/O, semaphores)

Artifacts to show (public-safe)

• raw scrape JSON (sources + timestamps)
• extracted events JSON
• final report (Markdown)
• any visuals/plots produced

Fill in real counts here:

• Releases processed: ___
• PDFs processed: ___
• Runtime: ___
• Cost per run: ___

Create account

Create account

Build narrative

Follow a coherent path from thesis to lab notes to proof-of-work instead of isolated pages.