About
I am a Computer Science undergraduate at the University of Colombo School of Computing with a deep passion for Natural Language Processing and Explainable AI (XAI). My work bridges the gap between rigorous research and scalable software engineering.
Currently, I am investigating explainability mechanisms in neural summarization models, specifically using Knowledge Graphs to validate factual consistency. I've also co-authored published research on SinhalaMMLU, a benchmark for evaluating LLMs on multitask language understanding for the Sinhala language.
When I'm not diving into transformer architectures or fine-tuning models, I enjoy building robust web applications using React, Spring Boot, and PostgreSQL. I believe that for AI to be truly impactful, it must be both efficient and understandable.
Experience
Research Intern · Language Technology Research Lab (LTRL), UCSC
Conducted comparative evaluation of LLMs on Sinhala educational tasks across 1,000+ questions. Co-authored SinhalaMMLU benchmark (ACL/EMNLP 2025). Investigated BERT-based architectures for low-resource languages.
- Python
- Transformers
- LLM Evaluation
- NLP
- SinhalaMMLU
Chairperson · IEEE Women in Engineering (WIE), UCSC
Led organizational activities, managed cross-functional teams, and supported event planning for the student branch affinity group.
- Leadership
- Team Management
- Project Coordination
Projects
MyStore — Inventory Management System
A full-stack system enabling real-time stock tracking, category management, and audit-friendly logs. Features JWT-based auth and a responsive dashboard.
- Next.js
- TypeScript
- Prisma
- MySQL
- TailwindCSS
SpotBiz — AI-Assisted Business Directory
Implemented a domain-specific RAG pipeline to study semantic retrieval on tag relevance. Features advanced filtering and distance-based sorting.
- Spring Boot
- FastAPI
- LangChain
- PostgreSQL
- React
Distance Calculator
Geospatial web app to calculate distances using the Haversine formula. Integrated interactive map visualizations with Leaflet.js.
- React
- TypeScript
- Zustand
- Leaflet.js
- OpenCage API
Research & Publications
SinhalaMMLU Benchmark
Co-authored a comprehensive benchmark for evaluating multitask language understanding in Sinhala, establishing baseline metrics for LLMs in low-resource settings.
- NLP
- LLM Benchmarking
- Low-Resource Languages
Explainable AI for Abstractive Text Summarization
Investigating factual consistency in neural summarizers using Knowledge Graphs. Developing a framework to trace information selection from source to summary.
- XAI
- Knowledge Graphs
- Transformers
- Fact Verification