My Recent Works

Evolution of Cooperation in LLM-Agent Societies: A Preliminary Study Using Different Punishment Strategies
Evolution of Cooperation in LLM-Agent Societies: A Preliminary Study Using Different Punishment Strategies

● Final year research project studying normative reasoning abilities of LLM agents in social dilemmas using the Smallville multi-agent simulation environment, under the supervision of Prof. Stephen Cranefield (Otago, NZ), Prof. Tony Savarimuthu (Otago, NZ), Dr. Surangika Ranathunga (Massey, NZ), and Dr. Nisansa de Silva (UoM, SL)

● Modified the Smallville environment to model and simulate the diner's dilemma scenario, including implementing the dilemma scenario logic, agent scheduling, developing and fine-tuning the logic for punishment mechanisms, implementing detailed dilemma decisions and reasoning logging

● Lead the paper writing by drafting the abstract, literature review, and conclusion.

● Accepted and presented at the International COINE Workshop (C-ranked), co-located with AAMAS 2025, Detroit, Michigan, USA — 20th May 2025

Technologies Used: Python, Tiled Map Editor, Vertex API, Groq API

Explainable Bus Arrival Time Prediction Model with Improved Features Related to Topography and Points of Interest
Explainable Bus Arrival Time Prediction Model with Improved Features Related to Topography and Points of Interest

● Accepted and presented at the 27th IEEE International Conference on Intelligent Transportation Systems (ITSC) 2024 in Edmonton, Canada (h-index 85)

● Improved bus arrival time predictions by incorporating topographical data (elevation, road bends) and Points of Interest (tourist attractions, healthcare, public spaces) for heterogeneous traffic conditions

● Automated extraction of 5 topographical and 12 POI-related features and built a dashboard for data visualisation

Technologies Used: Python, MySql, R Shiny

LangTrend — Tracking Language Representation in NLP Research
LangTrend — Tracking Language Representation in NLP Research

● Weekly automated pipeline that scans arXiv cs.CL submissions, detects explicitly mentioned human languages via full-text extraction (HTML → PDF → abstract fallback), and publishes an interactive dashboard tracking language representation over time

● Implements 6-tier language resource classification (Joshi et al. / Ranathunga & de Silva taxonomy), acronym conflict detection, and transparent false-positive flagging with links to source papers for manual verification

● GitHub Actions workflow runs every Tuesday; all manifest JSONs are committed to the repository as open data

Technologies Used: Python, arXiv API, Docling, Astro

Global Migration Trends: Visualizing the Movement of People (1990–2024)
Global Migration Trends: Visualizing the Movement of People (1990–2024)

● Interactive visual explainer of how global migration flows evolved from 1990 to 2024, built as a group project for CS5346 Information Visualization at NUS

● Interactive map of migration corridors that moves from continent-level patterns to country-level routes, with year and origin filters and supporting charts on shifting destinations and economic drivers

Technologies Used: React, TypeScript, deck.gl, MapLibre GL, D3

Model Calibration for Birth Delivery Mode Prediction: Caesarean Section Prediction
Model Calibration for Birth Delivery Mode Prediction: Caesarean Section Prediction

● Recreating and validating the model through calibration studies and employing uncertainty quantification to enhance predictive reliability in this critical medical domain

Technologies Used: Pandas, NumPy, SciPy, scikit-learn — Python

Churn Analyzer App
Churn Analyzer App

● Empowered management decision-making on employee retention by developing a customer churn analyzing app with key insights influencing employee attrition

Technologies Used: Pandas, Seaborn — Python, H2O Wave

E-Commerce Platform — DBMS Project
E-Commerce Platform — DBMS Project

● Created an online shopping website in comparison to Amazon, with a custom backend from scratch

Technologies Used: MySql, FastAPI — Python, Node, Vite — TypeScript

Employee Attrition Analysis Project
Employee Attrition Analysis Project

● Empowered management decision-making on employee retention at Marvelous Construction by identifying four key areas of insights influencing employee attrition

Technologies Used: Power BI, Tableau, Python (NumPy, Pandas, Scikit-learn, Seaborn)

Kickstarter Technology Projects Dashboard
Kickstarter Technology Projects Dashboard

● Developed an insightful dashboard by employing various data visualization techniques, condensing 16 technology subcategories into five informative charts

Technologies Used: Power BI, Tableau