About

I am a Machine Learning researcher working on Applied AI, Reinforcement Learning, Recommender Systems and Explainable AI.

I completed my PhD in Computer Science at Trinity College Dublin, studying user cold-start and reinforcement learning for recommender systems, and then worked as a post-doctoral researcher in the Distributed Systems Group at Trinity College Dublin, working on explainable multi-objective reinforcement learning.

Before my PhD, I spent over three years in industry as a Senior Machine Learning Engineer at Zone24x7, building production computer vision systems for retail analytics and security.

  • Machine Learning
  • Artificial Intelligence
  • Reinforcement Learning
  • Recommender Systems
  • Explainable AI
  • Computer Vision

Experience

Post-Doctoral Researcher — Distributed Systems Group

Apr 2025 – Jun 2026 CONNECT Research Centre, Trinity College Dublin · Dublin, Ireland

Doctoral Researcher

Apr 2021 – Mar 2025 Trinity College Dublin · Dublin, Ireland

Senior Machine Learning Engineer — Computer Vision & Video Analytics

Dec 2017 – Apr 2021 Zone24x7 Pvt Ltd · Nawala, Sri Lanka

Trainee Associate Electronics Engineer — Machine Learning

Nov 2016 – Feb 2017 Zone24x7 Pvt Ltd · Nawala, Sri Lanka

Education

PhD in Computer Science

2021 – 2025 Trinity College Dublin, The University of Dublin · Ireland

BSc (Hons) Electrical and Electronics Engineering

2014 – 2018 Sri Lanka Institute of Information Technology (SLIIT) · Sri Lanka
  • First Class Honours · GPA 3.82/4.00 (ranked 2nd)

Publications

Full list and citations on Google Scholar.

AcceptedTREX: Trajectory Explanations for Multi-Objective Reinforcement Learning
Dilina Rajapakse, Juan C. Rosero, Ivana Dusparic
XAI 2026 — 4th World Conference on Explainable Artificial Intelligence, 2026
PaperCode
Reassessing the Effectiveness of Reinforcement Learning based Recommender Systems for Sequential Recommendation
Dilina Chandika Rajapakse, Dietmar Jannach
SIGIR 2025 — 48th International ACM SIGIR Conference, 2025
PaperCodeOnline Material
RLT4Rec: Reinforcement Learning Transformer for User Cold Start and Item Recommendation
Dilina Chandika Rajapakse, Douglas Leith
arXiv preprint arXiv:2412.07403, 2024
PaperCode
A Good State Estimator Can Yield A Simple Recommender: A Reinforcement Learning Perspective
Dilina Chandika Rajapakse, Douglas Leith
OARS 2024 — 4th International Workshop on Online and Adaptive Recommender Systems, 2024
Paper
User Cold-Start Learning in Recommender Systems Using Monte Carlo Tree Search
Dilina Chandika Rajapakse, Douglas Leith
ACM Transactions on Recommender Systems (TORS), 2023
Paper
Fast and Accurate User Cold-Start Learning Using Monte Carlo Tree Search
Dilina Chandika Rajapakse, Douglas Leith
RecSys '22 — 16th ACM Conference on Recommender Systems, pp. 350–359, 2022
PaperCode

Research & Code

Code for my research is available on GitHub.

TREX

Trajectory-based explanations for multi-objective reinforcement learning agents.

PaperCode

RLT4Rec

A Transformer-based RL recommender for cold-start and established users.

PaperCode

MCTS Cold-Start Recommender

Monte-Carlo Tree Search for fast user cold-start learning, optimised in C++ for real-time use.

PaperCode

RL Recommender Benchmark

Reproducibility study of RL-based sequential recommenders against supervised baselines.

PaperCodeOnline Material

View all repositories →

Skills

Research Areas
Reinforcement Learning (MCTS, multi-objective RL, Transformer-based RL), Recommender Systems (sequential recommendation, user cold-start, offline & simulation-based evaluation), Explainable AI, Computer Vision (object detection & tracking, person re-identification, face recognition)
ML Frameworks
PyTorch, TensorFlow, Hugging Face Transformers, LLMs, OpenCV
Systems & Deployment
Docker, AWS, GCP, Kafka, Flask, OpenVINO, edge deployment (NVIDIA Jetson), Linux, Git, SQL, MongoDB
Languages
Python, C++/C, R