Can Koz

Experience

• 

  Nokia

  Jul 2018 - Aug 2018

  R&D Intern

  • Design UI by using Java for gateways which will soon enable wireless 5G networking in Turkey.• Design and implement UI that runs on Debian Kernel for ARM Devices.• Built UI using Java which supervises 2 million+ of nodes / gate-ways located around Turkey.• Implement controls to complete tasks faster for operating systems, software, and data.

• 

  TUBITAK UZAY

  Dec 2018 - Jan 2019

  Computer Vision Lab Intern

  • Learned the fundamentals of computer vision on Inria Holidays dataset.• Applied Histogram of Oriented Graphs and other feature descriptors using OpenCV on datasets provided by TUBITAK Space Technologies Research Institute.• Built a simple image search program which runs HOG Descriptor.• Used queries to make my image search program faster.• Successfully calculated the accuracy of my feature descriptor.

• 

  Oracle

  Jun 2019 - Aug 2019

  Software Engineer Intern

  I worked as a software engineer intern with Oracle Team. • Improved the run time of existing SQL Queries by 20% through code refactoring/rewriting.• Visualized existing data according to requirements.• Worked as an open source contributor to All Data Management Project which utilized Apache Hadoop.• Held, attended coding sessions on weekly online meetings and presentations.

• 

  Carnegie Mellon University

  Mar 2020 - Mar 2023

  Undergraduate Research Assistant

  • Implemented deep learning solutions to interdisciplinary problems in the Visual Design and Engineering Lab @ CMU.“Flaw Detection in Metal Additive Manufacturing Using Deep Learned Acoustic Features” NeurIPS 2020 • Built a CNN model that classifies flaws during the additive manufacturing process from audio samples.• Use transfer learning, deployment of a pretrained ResNet50 for sound classification, PyTorch, Python, and SciPy.• Visualization of results such as the confidence matrix, use of Adobe Photoshop for images on the paper.“Data Augmentation of Engineering Drawings for Data-Driven Component Segmentation” IDETC-CIE 2022 • Improved classification accuracy by 5% through feature engineering. I engineered Coefficient of Variation (COV) and implemented Zernike Moments.• Created a prepossessing step which combined Hough line detection with custom erosion and dilusion steps to segment different parts of the engineering drawings.• Built a PyTorch Geometric pipeline and implemented GraphSage model that reached 90% accuracy for classifying dimension / contour lines on engineering drawings. Show less

• 

  Stanford University

  Mar 2021 - Aug 2021

  • Taught Python programming language to 10+ students in the CS Bridge program. • Improved my teaching abilities by receiving feedback from Nick Troccoli. • As a volunteer section leader, I prepared and taught a weekly discussion section of 8-10 students to supplement professors' lectures.• Created and led weekly discussion sections for Code in Place, introductory programming class offered globally during COVID19 crisis.• Increased students' understanding of Computer Science concepts and fundamentals.

  • Section Leader - CS Bridge 2021

    Jun 2021 - Aug 2021

  • Section Leader - Code in Place 2021

    Mar 2021 - May 2021
• 

  Various Companies

  Jun 2022 - Aug 2023

  AI Consultant

  LLM Engineer• Fine tuning LLAMA 2 on a privately held dataset with the aim of automating brokers' tasks.• Text summarization• Named entity recognition• Joint pruning and quantization on LLMs (Alpaca, LLAMA 1 & 2) for performance benchmarking.• Deployment of LLAMA 2 based chatbot on Amazon AWS.Computer Vision Engineer• Spiking Neural Networks on event based video.• Neuromorphic Deep Learning and Event-based vision.• Simulation for AI models.