Doctoral Researcher · AI & Computing

Building trustworthy,
Efficient AI.

I research multi-agent reasoning, energy-efficient deep learning, and computer vision — with award-winning publications and accessible AI applications.

YA
Research collaborations & affiliations
About

Researcher in trustworthy & efficient AI

I am a fully funded Doctoral Researcher at Robert Gordon University, Aberdeen, developing the Global Ontological Reasoning Layer (GORL) for semantic interoperability in multi-agent conversational AI. My research focuses on trustworthy multi-agent systems, semantic reasoning, and AI reliability, with broader interests in energy-efficient deep learning, computer vision, and multimodal AI. I have also contributed to accessible AI applications. I have led first-author collaborations with researchers from Oxford, Newcastle, Stirling, Swansea, Surrey, and RGU, and have authored multiple peer-reviewed publications, including a Best Paper Award.

Experience

Research & professional path

2025 — Present

Doctoral Researcher · Robert Gordon University

Aberdeen, Scotland — fully funded studentship
  • Researching the Global Ontological Reasoning Layer (GORL) for semantic interoperability in multi-agent conversational AI.
  • Developing trustworthy multi-agent reasoning frameworks with structured knowledge representation, provenance, and verifiable agent communication.
2026 — Present

Visiting Lecturer · Informatics Institute of Technology (IIT)

  • Supervise undergraduate final-year research projects.
May 2025 — Present

Voluntary Researcher · Newcastle University

England — remote research collaboration
  • Working with Dr. Oktay Cetinkaya and Prof. Rishad Shafik on ML-driven food spoilage detection.
2020 — Present

Independent Researcher

  • Led first-author collaborations across Oxford, Newcastle, Stirling, Swansea, Surrey, and RGU.
  • Built TaproQA (agentic RAG QA) and EvoAgent (evolutionary optimization) for international challenges.
  • Created SSL400 and designed lightweight models via multi-objective optimization (NSGA-II, Optuna).
May 2023 — Aug 2025

Data Analytics & Engineer · Forestpin (Pvt) Ltd

  • Led development and deployment of the Forestpin Sanctions Screening Service.
  • Implemented phonetic name-matching algorithms over large datasets.
  • Built and optimized a Node.js/React API deployed on Azure (Linux).
Selected work

Projects & research

Each project has its own page with details, photos, conference videos, and links. Open any to explore.

PhD Research

Global Ontological Reasoning Layer

Semantic interoperability and verifiable communication for trustworthy multi-agent AI.

Multi-AgentKnowledge GraphsTrustworthy AI
2025 — PresentOpen →
★ Best RGU AI Thesis Award - 2025 Cohort

Human Action Recognition

Energy-efficient human action recognition pipeline and First large-scale Sri Lankan Sign Language dataset.

DatasetEnergy-efficientAccessible AI
2025Open →
★ Best Paper Award

FACTNet

A Flash-Attention-optimized CNN–Transformer framework for real-time video restoration.

Computer VisionTransformersReal-time
2025Open →
Challenge Entry · Top 8 Stages

TaproQA

Agentic retrieval-augmented question-answering framework selected for the IES GenAI Challenge comprising 650+ teams.

RAGAgentsGenAI
2025Open →
Challenge Entry

EvoAgent

Evolutionary optimization framework built for the SSBSE 2026 SBSE Challenge.

EvolutionaryOptimizationSBSE
2026Open →
Health · Vision

Brain Tumor Classification

Comparing evolutionary and Bayesian multi-objective optimization for energy-efficient brain tumor classification.

comparative studyDeep Learning
2024Open →
Health · Audio

Real-Time Snoring Analysis

A lightweight 1-D CNN for predicting snoring duration from audio signals to support patients with sleep apnea.

1-D CNNLightweightAudio
2023Open →
Computer Vision

Currency Authentication

Wearable, AI based framework for currency authentication and counterfeit prevention for the Visually Impaired community.

CNNClassificationSecurity
2023Open →

Let's collaborate

Open to research collaborations, talks, and supervision discussions.