DE.

Computer Vision Researcher

Dasith
Edirisinghe.

PhD Researcher at Queensland University of Technology

Brisbane, Australia

Computer Vision · Deep Learning · Representation Learning

View résumé Google Scholar
01 About

I am a PhD researcher in Computer Vision at the SAIVT Research Centre, Queensland University of Technology, Brisbane, Australia.

My research focuses on robust AI for autonomous perception, particularly 3D perception and representation learning. I am also interested in efficient deep learning, model optimization, and accelerating AI training and inference on modern hardware.

I hold a First Class Honours B.Sc. in Computer Science and Engineering from the University of Moratuwa, Sri Lanka.

Email GitHub LinkedIn ORCID
Latest Updates

News

Commenced doctoral research in Computer Vision at Queensland University of Technology, Australia.

Our chest X-ray report generation paper was published in IEEE Access.

Served as a peer reviewer for the MERCon Computer Vision track.

Our SpotKube paper was accepted at IEEE CloudCom 2024.

Began working as a Research Assistant at AITR Lab.

02 Education

Education

  • Ph.D. in Computer VisionQueensland University of Technology · SAIVT Research Centre
    Research focus: Robustness of 3D perception for autonomous systems
  • B.Sc. Engineering Honours in Computer Science and EngineeringUniversity of Moratuwa · First Class
03 Recognition

Honours & Awards

  • QUT Postgraduate Research Award (QUTPRA)Awarded the prestigious fully funded QUTPRA scholarship for PhD studies at Queensland University of Technology, School of Electrical Engineering and Robotics.
  • UTS President’s Scholarship (UTSP)Awarded the prestigious President’s Scholarship for PhD studies at the University of Technology Sydney, Faculty of Engineering and Information Technology.
  • Best Paper Nomination — Top 5IEEE International Conference on Cloud Computing Technology and Science (CloudCom), for “SpotKube: Cost-Optimal Microservices Deployment with Cluster Autoscaling and Spot Pricing.”
  • Dean’s ListAwarded for outstanding academic performance in Semesters 4, 5, and 8 at the Faculty of Engineering, University of Moratuwa.
  • Google Summer of CodeSelected by Google for its highly competitive open-source program. Contributed to Weaviate with a stipend and mentorship for the project “Make a New Weaviate Module.”
04 Publications

Peer-reviewed research

Selected publications

All on Scholar
2025
IEEE Access2025

Chest X-Ray Report Generation using Abnormality Guided Vision Language Model

Dasith Edirisinghe, W. Nimalsiri, M. Hennayake, D. Meedeniya, and G. Lim

IEEE Access, vol. 13, pp. 157651–157673.

Paper Code
2024
IEEE CloudCom2024

SpotKube: Cost-Optimal Microservices Deployment with Cluster Autoscaling and Spot Pricing

Dasith Edirisinghe, K. Rajapakse, P. Abeysinghe, and S. Rathnayake

Proceedings of IEEE CloudCom, pp. 87–94.

Paper Code
05 Experience

Research & industry

Experience

2024 — 2026

Artificial Intelligence Translational Research Lab · Sri Lanka · Part-time

Research Assistant — Machine Learning

Advised by Dr. Thanuja D. Ambegoda and Prof. Dulani Meedeniya.

Researched vision-language learning, 3D scene understanding, transparent-surface identification, and agentic AI for scientific research.

  • META-CXR: Developed a multi-encoder framework for chest radiograph understanding and report generation, published in IEEE Access.
  • DG-3DPlace and TSR-GS: Explored diffusion-guided 3D asset placement with Gaussian Splatting and self-supervised transparent-surface identification.
  • ScholarMind: Investigated literature-grounded agentic AI for research problem formulation and iterative reasoning.
Computer VisionVision-Language ModelsGaussian SplattingAgentic AI
2023 — 2026

Rabot Inc · Remote

Machine Learning Engineer

Developed production computer vision systems spanning vision-language models, metric learning, embedded inference, and robotic perception.

  • Improved binary visual-question-answering accuracy by 30% through LoRA fine-tuning.
  • Quantised YOLOv7 for Hailo hardware, achieving real-time 30 FPS inference.
  • Built a CUDA and TensorRT optimised ROS2 inference pipeline using ONNX Runtime.
PyTorchVLMsROS2CUDATensorRT
2022

Weaviate · Remote

Google Summer of Code Contributor

Created a text-summarisation module for search results and developed its inference service using FastAPI and Hugging Face Transformers.

Read final report
2021 — 2022

Sysco LABS · Sri Lanka

Software Engineering Intern

Helped modernise the Sysco Warehouse Management System by resolving React interface issues and improving Java service-layer performance.

06 Projects

A focused selection

Research projects

More on GitHub
01

Research project · University of Bern

Representation Learning for Image Order Prediction

Trained a sender–receiver system to encode and recover the order of image pairs through binary messages and learned embeddings.

Representation LearningPyTorchOptimisation
View project
02

Independent research

Contrastive Learning for ECG Signals

Compared SimSiam, BYOL, SWAV, and SimCLR for self-supervised representation learning on 1D ECG data.

Self-supervised LearningECG
03

Computer vision

GI Tract Anomaly Classification

Applied transfer learning with MobileNetV2 to classify anomalies in gastrointestinal endoscopy imagery.

CNNTransfer Learning
View code

Let’s connect

Interested in collaboration or have questions about my research? Please reach out!

udasith@gmail.com
GitHub LinkedIn Scholar ORCID