← All projects
Automationengineering projectPrototype

GD&T Drawing Tokenizer and Ballooning Prototype

A local web application for extracting PDF drawing text, mapping GD&T-related tokens, supporting manual correction, and exporting inspection-oriented data.

MY FOCUSDeveloped the full-stack prototype, token workflow, correction UI, and export paths.
FastAPIHTMLJavaScriptPDF.jsOCR libraries

ENGINEERING VALUE

Shows how software can reduce inspection-document preparation while preserving a human verification step.

PROJECT SCOPE

Heuristic drawing-tokenization prototype; OCR partly scaffolded

Explore the project checks ↗
ENGINEERING STUDY / DRAWINGJS—02

Engineering concept illustration.

01 / THE PROBLEM

The engineering challenge.

Inspection planning often requires repeated transcription of drawing callouts into structured reports.

02 / CONSTRAINTS

What the solution must consider.

  • Vector-text extraction is more reliable than scanned drawing OCR; font mapping and third-party license review remain necessary.

03 / ENGINEERING APPROACH

Reasoning before geometry.

01

Engineering method

Prefer deterministic vector text when present, provide explicit manual correction, and treat OCR as a separate lower-confidence route.

02

Workflow features

PDF page rendering; Vector text extraction; Token mapping; Manual correction; Balloon/report workflow; CSV/JSON/XLSX-style export.

04 / DESIGN & IMPLEMENTATION

Connecting intent to execution.

The FastAPI and browser workflow combines drawing-text processing, token organization and font mapping.

  • PDF page rendering
  • Vector text extraction
  • Token mapping
  • Manual correction
  • Balloon/report workflow
  • CSV/JSON/XLSX-style export

05 / VALIDATION

The checks behind the design.

Prototype

Human review remains part of the workflow. Extraction accuracy needs evaluation against representative vector drawings and scanned drawings; the OCR path is still partly scaffolded.

06 / OUTCOME

The value of the work.

Shows how software can reduce inspection-document preparation while preserving a human verification step.

THE ENGINEERING PRINCIPLE

Automate the workflow, including its failure cases.

Reliable engineering automation needs input checks, predictable output, clear diagnostics and a way to verify that quantities and references were preserved. Saving clicks alone is not sufficient evidence of correctness.

A question for a technical conversation

What would you test when the input is incomplete, duplicated or outside the tool’s supported scope?

Explore the engineering tools ↗

LET’S BUILD SOMETHING THAT WORKS

Good design belongs
on the shop floor.

Considering a role in mechanical design, manufacturing methods,
fixture design, or estimation? Let’s connect.