Case study
Faultrix: AI-Powered Construction Quality Control & Automated Report Generation
An AI-powered construction quality-control SaaS built solo from zero to production.
Problem
Site inspectors in Austria and Germany spend up to 4 hours daily compiling construction defect reports under ÖNORM B 2110 standards, facing slow turnaround, missing evidence trails, and disputed liability.
Role
Sole engineer and product builder: complete architecture, Next.js frontend, Convex backend, vision pipeline, Stripe billing, and DSGVO / ÖNORM compliance.
Architecture & Pipeline Summary
Next.jsConvexOpenAIClerkCloudflare R2StripeDockerPython
1. Photo Capture & Upload
Inspectors upload site photos with EXIF timestamps, GPS metadata, and inspection notes.
2. Multi-Modal Defect Analysis
OpenAI vision models analyze visual anomalies against construction defect taxonomies and ÖNORM standards.
3. Evidence Chain & Report Assembly
Generates ÖNORM B 2110-compliant audit reports in under 1 minute with SHA-256 evidence verification.
Approach
- Design a tight upload-to-report workflow around real construction site documentation workflows.
- Use OpenAI vision models with strict schema enforcement, structured output extraction, and cryptographic evidence hashes.
- Architect a serverless, compliant stack: Next.js, Convex, Clerk, Cloudflare R2, Stripe, Docker, and Python services.
Results
- Generated ÖNORM B 2110-compliant audit reports in under 1 minute from photo uploads.
- Achieved 98.4% industrial defect detection accuracy (threshold-dependent internal operational metric).
- Took an AI SaaS solo from concept to paying customer production with SHA-256 tamper-evident integrity.
Lessons
- User experience and workflow speed matter more to field inspectors than model size.
- Regulatory compliance (ÖNORM, DSGVO) and cryptographic evidence chains shape institutional trust.
- Shipping reliable AI requires rigorous fallback pathways for API rate limits and hallucinated outputs.