Orkun Temiz

An integration script between the Oracle ERP and third-party Primavera software.

I was the database and back-end developer of this project. I created the necessary web services and scripts integration using PL/SQL and XML. Also, I built Tableau dashboards to monitor real-time integration status.

An integration tool that integrates Oracle ERP and third-party master data-management software.

I was the back-end developer of this project. I was responsible for developing the web services required for the integration and PL/SQL scripts, which trigger necessary events in the Oracle database.

A Python-based script for claim extraction and verification of given claims, using Wikipedia articles as a source. Given a factual claim involving one or more entities, the script must extract evidence from sentences, table cells, table captions, and list items supporting or refuting the claim. Using this evidence, it labels the claim as "supported," "refuted given the evidence," or "not enough info (NEI)" if there isn't sufficient evidence to either support or refute it.

I developed a machine-learning model for text extraction and deep learning models to verify extracted texts. To obtain the ground truth about the claim and query enhancement, I integrated it with NoSQL and relational databases used via query requests sent from Python.

A Tableau report set developed for production planning and scheduling in a job shop-based environment.

I worked both as a data analyst and database developer in this project. I created necessary raw data using SQL queries from an Oracle database and developed Tableau reports for data visualization and data storytelling.

An Excel VBA script for material-requirement planning considers a project's supply and demand information. It then allocates the current supply to the demand by considering the priority and demand's timeline. The script also recommends additional supply if there is a supply shortage. The supply and demand information is retrieved via SQL queries from the Oracle ASCP module.

K-nearest neighbor algorithm is an algorithm widely used for data classification. However, most of this algorithm's implementation runs on the CPU, which causes performance loss in high volume data. The parallel programming implementation of the KNN algorithm utilizes the GPU cores and gives the same output one hundred times faster than CPU implementation.

I developed the CUDA code with a C++ platform using Visual Studio as the IDE.

This project aimed to efficiently classify a high volume of data, and the algorithm is used in a medical document classification script developed in Python.

A zero-shot fact extraction and verification pipeline to verify user tweets related to COVID-19 against the medical articles. The pipeline composes preprocessing of user posts, claim extraction, document retrieval, evidence selection, and verdict assignment components. The pipeline does not need to see previously labeled posts, unlike numerous supervised studies in the literature. Instead, it uses the zero-shot capabilities of existing models. The system is originally Python-based. However, modules written in C++ and CUDA are imported to Python and integrated into this project. This pipeline uses state-of-the-art NLP models and text analytics methods. For the medical articles, the Kaggle dataset (https://www.kaggle.com/allen-institute-for-ai/CORD-19-research-challenge) and medical archives Pubmed and Medrixv are used.

Python-based scripts which do automatic crypto trading in well-known crypto platforms. The scripts check on websites, Twitter, and certain mathematical metrics, and then according to the triggering conditions, the scripts take automatic actions in the crypto platform. The GitHub repository is currently private. However, I made it available to the public in the upcoming days.

A deep-learning model trained on the FER-2013 dataset contains 35,877 labeled 48x48 images belonging to seven categories of emotions. Using Python scripting language, an image classification architecture is developed on the TensorFlow library.

A Java object-oriented programming implementation of a chess game. Currently, the application does not have GUI; instead, the chess table and pieces are printed as strings to the application's console.

I am presently adding a suitable GUI.