Quartech: Automating Deployment for an AI Chatbot Frontend
About Quartech
Quartech is an IT services company based in Burnaby, BC with over 150 employees and contractors placed at various clients across Canada and the US, it is a leading supplier of bespoke technical solutions for both public and private sector organizations. Kai worked at Quartech from 2019 to 2025 with a number of clients including Telus, the Royal British Columbia Museum, Orange County Police Department, British Columbia Ministry of Post-Secondary Education and Skills, Ministry of Social Development and Poverty Reduction, and the Office of the Attorney General.
The Problem
Quartech developers were wanting to make a web page front-end to the Azure OpenAI service which would be able to query a library of documentation and provide natural language answers via the use of a large-language model.
The program required several specific packages deployed by the node package manager as well as a number of specific external dependencies, additionally the team wanted a way to deploy this in a repeatable fashion and at scale with minimal effort when pushing code changes.
The Solution
Kai adapted the developer’s solution to work inside a docker container, building from the ground up a dockerfile to include all needed dependencies and to execute the developer’s code. Through using a git repository in Azure DevOps and a DevOps pipeline, this allowed the team the ability to push the solution to an Azure Container Registry. Through using the Azure container registry the solution could easily be deployed into either Azure Container Instances or the Azure Kubernetes Service.
The Result
The end solution reduced build times significantly, taking a manual process which took a significant amount of time into an automated process which could take code-updates and push them to an updated container image in a matter of minutes. This provided developers the ability to easily test out new features and observe how their application behaved in an environment which would mirror that of a hypothetical production environment, making the go-to-live significantly easier once they completed the application.