Problem Statement
Our business aims to provide a globally accessible benefits management platform. The users of the platform that configure benefits and employee data management are increasingly facing the challenge of using the platform effectively since they are from diverse audiences across different languages and regions. The growing demand for fast, accurate, and scalable translation solutions led to the need for an automated, AI-driven text content translation tool.
Solution Summary
To address these challenges, I contributed to the development of the user interface for the tool. This tool uses artificial intelligence to automate the translation of text content across multiple languages for localization.
It integrates into our existing content management systems, enabling businesses to translate content in real-time or schedule translations as when they are required.
The solution also includes user-friendly interfaces for non-technical users to easily request translations and review output.
My Role
I was responsible for building the user interface (UI) in Angular, ensuring a smooth and intuitive experience for users managing translations. This included creating features for selecting target languages, and viewing translation progress through a clear and descriptive user experience with the Angular Material SDK and the Microsoft SignalR framework for reading asynchronous percentage value events from the API. I spent time researching and discussing the API implementation with the senior developers on my team.
Technologies Used
- Angular
- C#
- MongoDB
- MicrosoftSQL
- OpenAI API
Conclusion
This was a successful project which translated into investment into AI Language translation support projects at Darwin. AI is highly efficient at producing translations for large amounts of text content, however it is not able to fully maintain the intent, tone, and cultural relevance of the original content. We had to adapt via custom prompts that could be submitted to the LLM for overriding the erroneous content and this may need to be further enhanced via the LLM’s role configuration and custom context configuration.