Ibrahim Salman
Send a brief
Ibrahim Salman
Send a brief

Guides & Benchmarks

Technical guides for evaluating RAG assistants, fixing reading order in scanned-book OCR, and building fail-closed monitoring. The OCR benchmark includes measured results and downloadable data. The RAG evaluation guide proposes a protocol; it does not report a completed UET GPT accuracy study.

  1. Scanned-Book OCR: Gutter and Reading Order

    Stop OCR from interleaving facing pages with gutter splitting, reading-order checks, and measured Tesseract vs RapidOCR tradeoffs.

    Document Intelligence

  2. RAG Evaluation: A 30-Question Scorecard

    Build a small human-checked test set and score retrieval, citation support, refusal correctness and latency.

    AI & RAG

  3. Fail-Closed Job Board Monitoring

    Build resilient web scrapers that report broken targets instead of silently failing, using SQLite WAL mode and bounded retries.

    Automation & Monitoring

  4. Scanned-Book OCR Benchmark

    Compare 14 scanned spreads across Tesseract, RapidOCR, and B.L.A.S.T. with per-page CER, WER, reading order, timing, and downloadable JSON.

    Benchmark & Dataset

Interactive Tools

Compare cloud OCR API pricing (AWS Textract, Google Cloud Vision) against self-hosted CPU pipelines for your expected monthly document volume.

Launch OCR Cost Estimator →