🧠 Low-Resource Cross-Architecture Knowledge Transfer
Vision-Language Models work done during internship at SMU & M3S. Focus on transferring knowledge between architectures without paired supervision.
PyTorch
Explainable AI
VLMs
SASIKA
AMARASINGHE
Final-year Electronic & Telecommunication Engineering undergraduate at University of Moratuwa.
Focusing on Spatio-Temporal 3D Visual Grounding on Dynamic Point Clouds. Passionate about 3D Computer Vision, Multimodal Learning, and Computational Photography.
Researcher / Engineer / Developer
I have studied / worked at
M3S LAB
ENTC
Experience
Research Intern (Remote)
Jul 2025 - PresentA*STAR Institute of High Performance Computing (IHPC), 🇸🇬 Singapore
- Conducting research on Vision-Language Models for low-resource environments.
- Developing cross-architecture knowledge transfer pipelines using PyTorch.
- Collaborating with international research teams on cutting-edge AI solutions.
Research Intern
Dec 2024 - Jun 2025Singapore Management University & M3S, SMART Lab, 🇸🇬 Singapore
- Developed 3D visual grounding frameworks for dynamic environments.
- Implemented real-time inference models on edge devices like Jetson Orin Nano.
- Authored technical reports and contributed to lab publications on multimodal learning.
Team Lead - QSentinel
Nov 2024IEEE ComSoc Student Competition 2024
- Received Honorary Mention for innovative communications solution.
- Led a team of 4 engineers in designing a robust signal processing prototype.
- Presented technical findings to a panel of industry experts and academics.
Team Lead - EcoNova
Sep 2024Spark Challenge 2024, UoM
- First Place Winner in the Electronic and Telecommunication Engineering Department challenge.
- Designed sustainable IoT monitoring systems for agricultural use cases.
- Optimized power consumption for long-term remote deployment.
Writing
-
Capture Activations and Attention Maps using PyTorch Hooks
Technical Guide • 5 min read
-
A guide for the unknown: #ScholarX
Career & Mentorship • 8 min read
-
Enhancing Low Light Images: A Deep Dive into Autoencoders
Computer Vision • 12 min read
-
Hybrid Images through Frequency Wizardry
Image Processing • 6 min read
Selected
Tools & Work
🦾 Bin Picking Robot Arm System
Computer Vision System for Bin Picking Task using image Segmentation models such as SAM, DeepLab, Unet, and Segnet on an industrial robot arm.
PyTorch
Edge Comp.
🔆 D2BGAN
PyTorch Implementation of paper titled D2BGAN: A Dark to Bright Image Conversion Model for Quality Enhancement without Paired Supervision.
PyTorch
CycleGAN
🌌 t-SNE visualizer
Converts any image dataset into 2D t-SNE visualization for interactive 3D mapping of photography based on visual features.
Image Proc.
Viz
🔊 MFCC Speaker Recognition
Audio Signal Processing Challenge 2024 on Robovox Dataset for far-field speaker recognition by a mobile robot.
PyTorch
Audio
👁️ Vision Encoder Eval Tool
A tool to evaluate different explainability methods for Vision Transformer Architectures.
LLMs
Hugging Face
🎥 Real-Time 4K Video Exposure
Extended WACV 2024 implementation for real-time 4K video exposure correction pipeline.
Python
OpenCV
🌐 Jetson-Hololens Integration
A distributed system integrating NVIDIA Jetson for edge computing with Microsoft Hololens for AR visualization.
AR/VR
Networking
🚶 Infra-Free 3D Human Tracking
Implementation of Kalman Filter for tracking humans in 3D point cloud data without pre-installed infrastructure.
Kalman
Point Clouds
iCliQ
💡 iCliQ Wearable
A wearable clicker for public speakers and other professionals. Custom PCB and enclosure design.
PCB
ESP32
🤖 Social Media Bots
Automated bots for Instagram and Twitter using Selenium and Python for engagement automation.
Python
Selenium
🔌 Transformer Web App
A comprehensive web application for managing power transformer inspections with AI-powered analysis.
Springboot
React
Skills &
Technologies
Core Stack
Python
C++
PyTorch
Docker
ROS
Linux
Git
OpenCV
Arduino
SQL
Development
React
JavaScript
HTML5
CSS3
Java