Back to Projects
Trilingual Multimodal Financial Support Ticket Classification System
A comprehensive group data science and software engineering research project aimed at building a robust multilingual financial support ticket classification system. It combines modern NLP techniques with OCR to understand context from both typed text and attached images, ensuring highly accurate ticket routing in a complex linguistic environment like Sri Lanka.
Project Overview
Project Type
Group Project
Year
2026
Focus
Machine Learning / Data Engineering
My Role
Developer
Technologies Used
React
Node.js
PostgreSQL
Docker
Hugging Face Transformers
My Contributions
- Architected an end-to-end intent classification pipeline supporting 5 languages (English, Sinhala, Tamil, Singlish, Tanglish).
- Integrated OCR-derived text from receipts/screenshots with standard ticket text for multimodal context.
- Engineered real-time data streaming and deployed the model using Docker and Hugging Face infrastructure.
- Managed version control and CI/CD pipelines across the team.