Research and publications

Trustworthy multi-agent reasoning, energy-efficient deep learning and AI that widens access, with the peer-reviewed papers behind it.

PhD research, 2025 to present

Global Ontological Reasoning Layer

When agents built by different teams talk, the same word can mean different things. GORL lets agents detect that mismatch in dialogue, negotiate a shared meaning and keep a verifiable record of what was agreed, so multi-agent systems stay reliable and aligned with people.

About the project

Retail agentShip order 4471 to the customer's address.

Logistics agentI hold two kinds of address: billing and delivery. Which one do you mean?

Retail agentDelivery. In my schema that is shipTo.

Agreed: shipTo maps to deliveryAddress. Recorded for both agents.

Best RGU AI Thesis Award, 2025 cohort

Energy-efficient human action recognition and SSL400

An end-to-end sign language recognition pipeline tuned for low energy use, and SSL400, the first large-scale Sri Lankan Sign Language dataset.

Procedia Computer Science, 2025

 Best Paper Award

FACTNet

A Flash-Attention-optimised CNN and Transformer framework that restores video in real time.

9th SLAAI International Conference on AI, 2025

More projects

Publications

  • FACTNet: A Flash-Attention-Optimized CNN Transformer Framework for Real-Time Video Restoration
    Y. Abhishek, N. D. Kodikara. SLAAI ICAI. Best Paper Award.
    Read the paper
    2025
  • End-to-End Sign Language Recognition Pipeline: Towards Energy-Efficient Modeling
    Y. Abhishek, D. K. Sumanathilaka. Procedia Computer Science.
    Read the paper
    2025
  • SSL400: A Comprehensive Word-Level Dataset for Sinhala Sign Language Recognition
    Y. Abhishek, D. K. Sumanathilaka.
    Read the paperDataset
    2025
  • Deep Learning in Sinhala Sign Language Recognition: A Systematic Decade-Review
    Y. Abhishek, D. K. Sumanathilaka.
    Read the paper
    2024
  • Lightweight 1-D CNN Model for Predicting Snoring Duration
    Y. Abhishek, D. K. Sumanathilaka.
    Read the paper
    2023
  • Deep Learning Framework for Currency Authentication and Counterfeit Prevention
    Y. Abhishek, D. K. Sumanathilaka, et al.
    Read the paper
    2023