Jon Karakus

Machine Learning Researcher

Smith Engineering- AMSPLaboratory | Queen’s University•  March 2024 - Present•  Kingston, ON

• Researching supervised learning and data mining (Python) for audio datasets. Submitted research paper to IEEE. • Implemented and optimized classifiers for emotion detection using Scikit, achieving ~80% f-score (across 4 emotions). • Developed high/low pass filters and LMS noise cancellation algorithm to prepare data for feature extraction. • Developing a 10-layer convolution neural network (CNN) using Keras, to surpass the ~80% f-score obtained by SVM. • Awarded 2nd place and a scholarship prize among 60 competing undergraduate research projects

Hardware Data Specialist

Contextual AI•  July 2024 - Present•  San Fransico, CA (remote)

Generated in-depth evaluation reports for Large Language Models (LLMs), identifying hallucinations, inaccuracies, and document retrieval performance for semiconductor clients such as Qualcomm. Architected hardware component datasets to optimize training for Retrieval-Augmented Generation (RAG 2.0) model.

Engineering Specialist Intern

Thermo Fisher Scientific•  May 2023 - August 2023•  Toronto, ON

Wrote and executed test protocols for ~$2 million of pharmaceutical processing and automated packaging robots. Implemented databases (SQL) to collect pharmaceutical machine data automatically. Organized 350+ engineering tasks and monitored team capacity metrics using PowerBI.

Embedded Systems Engineer Intern

WorldStarTech.•  May 2022 - August 2022•  Toronto, ON

Designed a portable spectrometer instrument using a Hamamatsu C12880MA module for manufacturing and QC. Utilized Object Oriented Programming (C++) to control components. Designed and produced a 2-layer PCB using KiCAD.

Embedded Software Engineer Intern

WorldStarTech.•  May 2021 - August 2021•  Toronto, ON

• Engineered an intelligent UV curing system in C# with a Meca500 6-axis robotic arm for mounting optical lenses and motors. Created a 3D-printed gripper and UV holder, optimizing the arm's versatility and assembly efficiency. • Automated manufacturing assembly line, yielding a $47,000 monthly revenue increase through higher part quantities