Vithursan Thangarasa

Machine Learning Researcher/Practitioner

account_box About me

My name is Vithursan Thangarasa, friends and family call me Vithu. I am working towards an MASc under the supervision of Dr. Graham W. Taylor at University Guelph's Machine Learning Research Group (MLRG). I am interested in developing Continual/Lifelong Learning algorithms for Deep Neural Networks and exploiting high performance computing (HPC) technologies for large-scale machine learning applications.

I completed my BEng in Engineering Systems and Computing at the University of Guelph. Here, I focused on designing integrated computer based engineering systems which included software development, computer hardware design, mechanics and energy transfer, signal processing and digital process control.

perm_phone_msg Contact me

pin_dropToronto, Canada

stars My Skills
Machine Learning
Software Development
Embedded Systems

card_travel Work Experience

Machine Learning Engineer, Computer Vision

placeTesla, Inc. today May 2018 - Present

Working with the Advanced Technology Group (ATG) on confidential future products and computer vision systems for Tesla's vehicles

Graduate Research Assistant

placeUniversity of Guelph today May 2017 - Present

Research focuses on developing lifelong learning, meta-learning and curriculum learning algorithms for Deep Neural Network models to autonomously learn online over continuous streams of non-stationary data.

Deep Learning Data Scientist

place Scotiabank, Artificial Intelligence and Machine Learning Group today Sept 2016 - Dec 2016

Worked on a AI-powered financial chatbot for Scotiabank where, I mainly focused on Natural Language Processing/Understanding, Conversational Modeling, TensorFlow, Docker, AWS, and followed the Scrum methodology.

Hardware and Systems Developer

placeON Semiconductor today May 2016 - Aug 2016

During my time at ON Semi, I worked on embedded software within the Hardware and Systems team using Scrum methodology, and mainly focused on C, ARM Cortex-M4, BLE 5.0, Eclipse for the company's new ultra-low-power multi-protocol BLE 5.0 SoC: RSL10.

Software Engineer, Video Compression

placeEvertz Microsystems Ltd. - Head Office today Jan 2015 - Aug 2015

As a software engineer working on video compression, I worked on multiple projects on the H.264 and H.265 (HEVC) encoder/decoder software library for the company's main Software Defined Accelerated Encoding Platform: 3480TXE.

Mobile Application Developer, Android

placeJamdeo Ltd. (Flextronics & HiSense Joint Venture) today May 2014 - Aug 2014

I worked with the Advanced Architecture team to develop security libraries for machine-to-machine communication on an Internet-of-Things (IoT) application, where I used C, Eclipse, and Android NDK.

school Education

MASc, Machine Learning and Artificial Intelligence

place Machine Learning Research Group, University of Guelph today May 2017 - Present

Advisor: Dr. Graham W. Taylor

BEng, Bachelor of Engineering (Honours, Co-op)

place University of Guelph today Sept 2012 - Apr 2017


Magnet Loss in PyTorchmore_vert

work 127 work 11

PyTorch implementation of the Magnet Loss for Deep Metric Learning, based on the paper Metric Learning with Adaptive Density Discrimination by Oren Rippel et al. from Facebook AI Research (FAIR) that was accepted into ICLR 2016.

Terraform + AWSmore_vert
terraform-aws-spotgpu close

work 92 work 11

A Terraform module for provisioning EC2-based Spot Instances on AWS, specifically for Deep Learning workloads on Amazon's GPU instances, by taking advantage of automation and friendly declarative configurations.