Australian Moulding & Door Company (AMDC) is an Australian manufacturing business specialising in bespoke timber mouldings, including architraves and skirting boards designed to match historical or legacy architectural profiles. The company’s products are frequently used in renovation and restoration projects where homeowners, builders, or architects require mouldings that replicate older timber profiles no longer produced by large-scale manufacturers.

AMDC maintains an extensive catalogue of moulding patterns developed over many years of custom manufacturing. However, identifying the correct pattern for a customer request traditionally required a manual process. Customers would typically send photographs or physical samples of mouldings to AMDC, where staff would use flatbed scanners or visual inspection to capture the shape profile.

The resulting silhouette would then be manually compared against a printed catalogue of profiles in order to identify the closest match. This process was labour intensive and highly dependent on individual expertise. It also limited the organisation’s ability to respond quickly to customer enquiries and made it difficult to service customers across Australia without receiving physical samples.

AMDC sought to modernise this process using artificial intelligence to automate pattern recognition and dramatically improve response times.

The Full IDP Solution phase, guided by the learnings from the POC, delivered a full-scale production-ready solution with enhanced capabilities:

Key Features of the solution

• The solution was built using the following AWS services:

  • Amazon Bedrock – Generative AI reasoning and pattern interpretation

  • Amazon S3 – secure storage of uploaded moulding images and silhouette data

  • AWS Lambda – serverless image preprocessing and vector generation

  • Amazon API Gateway – secure API interface for the web application

  • Vector similarity search (via AWS-managed services) – rapid matching of moulding vectors against catalogue profiles

  This architecture provides a scalable, cost-efficient AI platform capable of supporting increasing volumes of customer requests.

The legacy moulding identification process presented several operational challenges.

First, the identification process relied heavily on manual inspection. Staff were required to visually compare scanned silhouettes or photographs of mouldings against a physical catalogue of previously manufactured profiles. This process could take significant time and required specialised knowledge to identify subtle differences between similar patterns.

Second, the accuracy of the process depended heavily on the quality of input material. Customer photographs varied widely in lighting conditions, perspective, and scale. In many cases, customers were required to physically mail timber samples to AMDC so that accurate scans could be captured. This introduced delays and created logistical barriers for customers located outside AMDC’s immediate region.

Third, the manual approach limited scalability. As customer enquiries increased, the business faced growing operational overhead in reviewing and identifying moulding profiles. The company needed a solution that could automate silhouette extraction, analyse geometric features, and compare shapes against their existing catalogue.

AMDC engaged Mantalus to design an AI-driven solution that could analyse uploaded images, generate vector representations of moulding profiles, and use Generative AI to identify the closest matching patterns within their catalogue.