Yokhesh Krishnasamy Tamilselvam

Machine Learning Engineer

Followers

Following

3-5 yearsLondon, ON, Canada

Machine Learning Engineer

Mid-level3-5 yrs

$48/hr

PythonMachine learningComputer VisionNatural Language ProcessingC++GithubML frameworksML libraries Interpersonal SkillsCommunication skills

Anna University

Electronics and Instrumentation Engineering

Bachelor's DegreeClass of 2016

Clemson University

Electrical and Computer Engineering

Master's DegreeClass of 2018

University of Western Ontario

Robotics and Machine Learning

Doctoral DegreeClass of 2023

Human Centered Lab

full time contract10/2017 - 1/2018

Designed an RNN-based Natural Language Processing model in Python to classify the emotions of over 1500 patients to enhance the efficacy and provide a more individualized treatment based on the patient's emotions
Scraped the text data from over 3000 web posts provided on the "National Association for Continence" website to be used as the training and validation dataset for the NLP model
Performed data pre-processing steps and visualization procedures using SAS Text Miner and Linguistic Inquiry & Word Count software to prepare the dataset for analysis

Ph.D. Researcher

full time contract1/2019 - 6/2023

Organized and guided a research team to design robotic and machine learning techniques to enhance the efficacy and accuracy of diagnostic and treatment procedures for Parkinson's Disease (PD)
Developed an AI-enabled medical robot that can perform an automated and adaptive patient assessment using virtual reality programs
Designed a mobile robot system in ROS and a deep learning model using Keras that can learn and imitate human actions based on the sensorimotor optimization principles to automate the assessment process carried out by the medical robot
Built a supplementary deep neural network model that uses features obtained from the medical robot to diagnose PD at a much earlier stage than currently possible, at an accuracy of over 85%, thereby providing treatments in a timely manner

Singapore University of Technology and Design

full time contract3/2018 - 12/2018

Developed a vertical climbing modular robot equipped with an on-board visual support system for cleaning the floors and windows of high-rise buildings
Designed a lightweight Deep CNN (DCNN) model that can detect litter at an accuracy of 96 % and works 50% quicker than the SSD Mobile Net and SSD Inception object detection frameworks
Implemented a real-time path planning algorithm based on the DCNN’s output using Depth-First Search and Probabilistic Road Map techniques to automate the tasks of the cleaning robot, thereby reducing the cleaning time significantly
Designed a secondary CNN model to detect cracked glasses at an accuracy of 87% and incorporated the model into the robot to avoid cleaning the cracked regions of the windows to ensure the structural stability of the windows is not compromised

Singapore University of Technology and Design

full time contract4/2017 - 10/2017

Led a team of four researchers in collaboration with the Massachusetts Institute of Technology (MIT) to implement an automated fault diagnosis mechanism in a bio-inspired re-configurable robot using Support Vector Machine (SVM) and neural networks
Designed a machine learning algorithm that can identify locomotion faults in real-time at an accuracy of 90% using the data from the on-board IMU sensor, thereby informing the robot to reconfigure its morphology to overcome the fault
Tested the fault detection model, which indicated that the fault was identified within 20 seconds from the commencement of the robot operation, and published the results in a high-impact journal