New solution provides Great Vantage with solid foundation to build its GenAI capabilities

Database of 40,000 action items supported development of predictive answers

Great Vantage, based in Malaysia, is the developer of ehs tools, a suite of software as a service (SaaS) tools that help to streamline complex workflows to ensure seamless collaboration across various sectors, with oil and gas as a key market. The company was founded by engineers backed up by more than 30 years’ experience in engineering and operations. Its solutions are purpose-built to track actions for process safety and risk assessment methodologies, such as hazard and operability study (HAZOP) and hazard identification (HAZID), which are used by multiple industries to support safety measures and manage risk. 

The company’s flagship Actions Tracker SaaS tool (ehs tools) makes it easier for customers to address operational challenges for engineering, procurement, construction, installation and commissioning (EPCIC) project delivery guidelines and minimise the risk associated with tracking actions manually, such as with Excel files or email notifications. Manual tracking is time consuming and comes with a greater risk of costly mistakes and delays in performing remediating actions.  

The EHS Tools Activity Tracker, running on AWS Cloud, manages mass actions in EPCIC projects, as well as operations. It also helps in managing submissions for green building certifications. Managing risk in this way helps Great Vantage’s customers prevent major accident event or avoid expensive multi-million-dollar mistakes.

Great Vantage operates with a principle of continuous improvement for its products, implementing new AI services using AWS technologies. Using its database of over 40,000 action items it used GenAI to provide predictive actionable advice.

Building a solid infrastructure foundation to provide GenAI services

SoftwareOne provided consulting advice on how the infrastructure should be designed, including which AWS services would provide an optimal architecture and which database type it could use to feed vectorised data into the AI system. SoftwareOne recommended Amazon OpenSearch Serverless as the vector database for this project, to ensure low operational overhead while maintaining performance at scale.

SoftwareOne helped the Great Vantage IT Team build its AI knowledge base as well as set up the GenAI part of the infrastructure. The project, which had a delivery window of two months, uses AWS Bedrock, a fully managed service that supports the AI foundation models Anthropic Claude 3.5 Sonnet, and Amazon OpenSearch Service as its vector database because of the service’s ability to deliver fast and effective search capabilities. Amazon S3 serves as the centralised storage for historical ticket data and technical PDFs. AWS Lambda acts as the orchestration layer, processing user queries, invoking Bedrock to generate responses, etc. Amazon API Gateway provides the user-facing endpoint for query input and response delivery.

The company’s main reason for developing the solution was to look at how feasible it would be to incorporate AI into its flagship tool. As Hui Siang Yeo, business development manager and project lead at Great Vantage explains: “We needed to find out if AI could do the job,” he says. “Next, we needed to understand the business benefits with the AI route. And, based on the cost, could we monetise the functionality as an additional module and a new charging model?”

In projects of this nature, the scope of work is usually pre-determined. For Great Vantage, one of the advantages of working with SoftwareOne was the flexibility it allowed in the development process. “We were learning what was possible as we went along,” says Siang Yeo. “You can’t always shift things around in a project like this as it uses up too many resources.” 

SoftwareOne worked with Great Vantage to thoroughly test the system and check the accuracy of the answers provided by the GenAI solution. Checks included looking at search accuracy, keyword tolerance (ie. the ability of GenAI to deliver accurate results where typos are input) and speed (how quickly the suggestions are generated). This has enabled the system to go live while refining it as part of ongoing development.