LingoLah Malaysian Slang Translator is a web-based AI translation system developed to process Malaysian code-switched and slang-rich text. The system supports mixed-language input involving English, Bahasa Malaysia and Chinese, and translates the input into a single selected target language.

The system uses a React-based frontend connected to a FastAPI backend, with a fine-tuned Qwen3.5 9B large language model running locally through Ollama as the main translation engine. Before translation, the input is processed using token-level language identification to recognize the language of individual words or segments within a code-switched sentence. The processed text is then sent to the AI model, which generates the final translation in English, Bahasa Malaysia or Chinese.

In addition to translation, LingoLah provides an AI Dictionary for Malaysian slang and colloquial expressions, file text extraction for PDF, DOC and TXT documents, translation history, and user profile management. These components are integrated into a single web platform to provide a practical solution for Malaysian multilingual communication.

The system was evaluated through functional testing and user acceptance testing, while the fine-tuned model achieved 87% translation accuracy in the reported project evaluation.