Personal · In progress

AI Document Processing API

Stop copying PDFs into downstream systems by hand — a B2B extraction API with async validation and confidence scores (personal product, in progress).

Role: Sole builder · personal product / API in progress

FastAPIOpenAIPythonAsync pipelines
POST /api/v1/extract
API preview

Drop in ai-document-1.jpg or a Swagger shot

pdf → structured
ai-document-2.jpg
validation
ai-document-3.jpg

Problem

B2B teams still push rates, invoices, and ops data through PDFs and spreadsheets. Downstream systems need structured, validated fields — not another manual copy-paste step.

Role

Personal project in progress: designing and building a FastAPI service for document extraction, async processing, and confidence-aware outputs that other systems can consume.

What I built

  • FastAPI service surface for submitting documents and retrieving structured results.
  • OpenAI-powered extraction from PDFs and spreadsheets into typed fields.
  • Async validation pipelines so long-running jobs do not block the request path.
  • Confidence scores so callers can gate automation vs. human review.
  • Patterns aligned with production Rate Sheet Agent work (Shipsoft) — reusable for B2B ops.

Stack

Python, FastAPI, OpenAI API, async job/validation pipelines.

Code

GitHub