Weld IT, a leader in welding process management, aimed to revolutionise how its customers handle critical welding documentation. The company is a Norwegian technology provider specialising in cloud-based quality management solutions for the welding industry. It serves a diverse customer base—spanning enterprise and SMB clients from specialised welding workshops to large offshore fabrication companies—and reported approximately 44,000 NOK per month in recurring revenue by early 2025. Most of its customers operate in regulated sectors such as oil & gas, marine, and construction, where welding documentation and compliance are essential.
Confronted with a major business challenge of manual, time-consuming, and error-prone data entry, Weld IT partnered with Cloud‑bridge to develop an automated solution. The project focused on harnessing advanced AI and serverless architecture to streamline the processing of Welding Procedure Qualification Records (WPQRs) and Welding Procedure Specifications (WPSs), ultimately enabling Weld IT to deliver significant operational efficiencies and scalability for its enterprise customers.
Challenge:
Weld IT’s customers were burdened by a manual, resource-intensive workflow for managing welding documentation. The process involved manually extracting key metadata from complex technical documents, such as WPQRs and WPSs, and then manually entering this information into their platform. This required specialised welding knowledge and was highly prone to errors.
The traditional method led to:
- High operational costs: Significant staff time was spent on data entry and quality control.
- Limited flexibility: The process was not easily scalable, hindering growth and making it difficult to onboard new, high-volume customers.
- Increased risk of non-compliance: Manual errors could lead to incorrect data, potentially affecting the integrity of welding procedures and compliance with industry standards.
Solution:
Cloud Bridge collaborated closely with Weld IT to design and implement a comprehensive, cloud-native solution on the AWS platform. The core of the solution was an automated, serverless pipeline that replaced the need for manual data entry.
The key components of the implementation included:
- Automated Document Parsing with GenAI: Using Amazon Bedrock with Anthropic Claude Sonnet 3.7, Cloud Bridge configured Large Language Models (LLMs) to parse and extract specific metadata from technical documents accurately. Prompts were engineered to pull out critical data points—such as weld types, materials, and test results—and to return structured JSON formatted to Weld IT’s schema. A deterministic rule engine layer was added to handle standards like ISO 15614 and NORSOK M 601 when the LLM could not fully automate calculations (as noted in the lessons learned).
- Serverless Ingestion and Processing: The solution was built using a serverless architecture based on AWS Lambda and Amazon S3. Documents uploaded by customers were stored in an S3 bucket, triggering a Lambda function that initiated the automated processing pipeline. The event-driven workflow eliminates the need for server management and scales automatically with demand.
- API-Driven Data Delivery: Once the data was parsed and validated, it was made available via secure APIs. The extracted data could auto-populate forms or automatically generate new WPSs from WPQR data, saving customers countless hours.
- Production-grade implementation: The live deployment generates roughly 23,200 NOK (≈ £ 18,500) per year in AWS spend, confirming that the solution is not a proof-of-concept but a revenue-generating, production-grade system (self-assessment PS004).
Benefits:
The solution delivered by Cloud Bridge provided Weld IT and its customers with tangible and quantifiable benefits, transforming their document management processes.
- Significant time savings: The automated system delivered an 80–90 % reduction in the time previously spent on manual data entry for WPQRs and WPSs. This allowed customers to reallocate skilled personnel to more strategic tasks.
- Improved accuracy and compliance: By automating data extraction and validation, the risk of human error was dramatically reduced. This enhanced the accuracy of the data and ensured better compliance with strict industry standards, building greater trust and reliability in Weld IT’s service.
- Enhanced scalability: The serverless architecture allows Weld IT to effortlessly scale its documentation processes to meet the demands of large enterprise customers.
- Immediate business value: The seamless integration with the Weld IT platform, using APIs, meant that customers could immediately benefit from the structured data.
Lessons Learned & Next Steps:
During implementation, a stretch objective was to automate WPS generation from WPQR data fully. In practice, manual adjustments were still necessary to tune parameters like material thickness and heat input. This shortfall stemmed from the complex interdependencies of welding standards, the variability of vendor documents and the limitations of the LLM with context-heavy calculations.
To address this, Weld IT is now:
- Combining GenAI inference with deterministic rule engines to handle standards compliance.
- Building a larger training library covering more vendor formats and edge cases.
- Exploring cost-optimised inference on EC2 for high-volume workloads.
These continuous improvement initiatives ensure that the solution evolves alongside customer needs and industry standards while maintaining Weld IT’s leadership in AI-driven welding documentation.
