Laserhub lifts quoting speed and pricing accuracy with Werk24’s engineering-grade PMI extraction
“The precision of our cost calculation strongly relies on Werk24’s PMI extraction — and we encounter almost no files where the tolerances of threads are not correctly extracted from technical drawings.”
— Christoph B. Roessner & Adrian Raidt, Co-Founders, LaserhubThe Challenge
Online manufacturers compete on speed, accuracy, and throughput. Laserhub’s engineers were spending 30–90 minutes per drawing interpreting PDFs, transcribing PMI, and normalizing units — time that should have gone into pricing strategy and customer value.
- High variability: every customer uses different title-block and PMI conventions
- Hidden risk: tight tolerances, special threads, or roughness calls could be missed under time pressure
- Scale limit: throughput was capped by headcount; quoting speed dipped during peaks
For a business built on instant, accurate quotes, manual reading was a structural bottleneck and a margin risk.
Why Now
- More drawings, more variability: growth amplified the parsing burden
- Customer expectations: fast quotes with precise, transparent cost drivers
- No appetite for template maintenance: solution must work on scans, not just perfect PDFs
The Werk24 Solution
Engineering-grade extraction for mechanical drawings
Werk24 reads 2D mechanical drawings (components & assemblies) — including acceptable scans and hand-drawn engineering drawings — and returns structured JSON with field-level confidence scores. We extract title-block metadata and feature data like dimensions, upper/lower tolerances, GD&T, threads, and surface roughness.
- Unit normalization and symbol mapping (DIN/ISO/ASME where applicable)
- Deterministic schemas or custom output that matches Laserhub’s pricing fields
- Quality gates to push accuracy up to 99% on critical features
Integration that fits the stack
Laserhub integrated via HTTPS webhooks and the Python SDK. Confidence thresholds auto-accept routine fields and route edge cases for review. No middleware was required: Werk24 shipped a custom schema mapping so payloads drop straight into the pricing engine.
- REST API + webhooks for lights-out processing
- Python SDK for pipelines and batch runs
- Data residency choice: EU or US; higher tiers support zero-retention
See it in action
Download assets
Implementation in days, not months
- Day 1–2: trial key, sample drawings, output schema defined
- Week 1: first workflow live (RFQ intake), confidence thresholds tuned
- Week 2+: rollout to pricing; optional quality gate for high-impact features
No labeled training data required — unless you request a custom quality gate.
Results
faster quotes
PMI extraction down to seconds per page.
extraction accuracy
Comparable to a trained external engineer; 99% with gates.
higher cost precision
Normalized tolerances & units reduce misreads.
throughput uplift
Scale quoting without adding headcount.
From developer: inside for developers: inside
Clean APIs, predictable JSON, confidences and webhooks. From PDF to production in hours instead of weeks.
pip install werk24 # Install Werk24
werk24 init # Obtain a trial license
werk24 health-check # Verify connection
Ready to see it on your drawings?
Try the online demo or request a trial API key (100 drawings).