Case study

Squint AI — Document-to-SOP Pipeline

An AI-powered document-processing platform that converts unstructured business documents into structured, editable standard operating procedures.

Snapshot

20+ SOP documents processedApproximately 50,000 words processedMore than 90% parsing accuracy

Technology

PythonGeminiFastAPIPydanticPyMuPDFCeleryAWS S3Next.js

Links

Case study

System breakdown

01

Overview

A document-processing platform for transforming messy business files into structured SOP data that can be edited and reused downstream.

02

Problem

Business process documents often arrive as PDFs, DOCX files, and plain text with inconsistent formatting, making them hard to convert into usable application data.

03

Solution

The pipeline extracts text and tables, normalizes document structure, and uses AI-assisted parsing to produce structured JSON.

04

Personal contribution

Built parsing workflows, structured output schemas, evaluation flows, and integrations across the document-processing stack.

05

Technical architecture

FastAPI services, Python document extraction libraries, Pydantic schemas, Celery jobs, AWS S3 storage, and a Next.js application layer.

06

Results

Processed more than 20 SOPs, approximately 50,000 words, and reached more than 90% parsing accuracy.

Visual layer

Screens, diagrams, or motion studies can drop in here later.

The page is already structured for future screenshots or architecture visuals without needing another layout rewrite.