Cahid Arda Öz Email & Phone Number

· Implemented modules within data processing pipelines to enhance efficiency and add new features using PySpark. Notable contributions include an outlier detection module, which improved the runtime significantly, a reporting module for accuracy and prediction distribution assesment via DataDog, and a module dedicated to calculating price features for products/stores.· Worked on developing a web application with Streamlit, focusing on explaining black-box machine learning forecasts using several Python libraries for AI explainability.· Impact: Reduced runtime of two complex algorithms from 4 hours to 0.5 hour on benchmark data. Integrated Dalex library to add machine learning model interpretability support to the MLFlow plugin of the internal forecasting library.· Used Python, MLFlow, Streamlit, Tensorflow, Sklearn, DataDog, Databricks, PySpark.

Istanbul, Turkey

· Worked on solving a niche image processing task that involved classifying objects in low contrast, microscopic images. Worked independently on researching, testing and evaluating different approaches.· Impact: Used MTRG Segmentation to create an annotated image dataset. Trained deep learning models such as CE-Net and Mask-RCNN using this data. Used the trained models and image processing methods to create a pipeline for segmenting objects in the niche dataset.· Used Python, OpenCV, Tensorflow, sklearn, AWS EC2 and S3

Istanbul, Turkey

· Learned about time series forecasting, focusing on demand forecasting for ATM units. Tested feature engineering, machine learning and deep learning methods. Used Keras to implement the TabTransformer, a deep learning architecture for modeling tabular data.· Impact: My TabTransformer implementation was added to the company’s collection of models.· Used Python, Tensorflow, Keras, sklearn and other Python libraries.

Istanbul, Turkey

Worked on the RE4DigiTR project, lead by Fatma Başak Aydemir. Helped develop ArTu, a model-driven web application for visualizing goal models based on a number of heuristics. · Worked on developing ArTu, a web application for visualizing requirements and user story sets as graphic goal models. Used JointJS library to improve the graphical interface. Added new features and improved user experience of the application. Deployed the application on Heroku. Used Audacity, OBS Studio and OpenShot Video Editor to record and compile a demo video demonstrating the ArTu application.· Impact: Significantly improved user experience and published a paper about the project. Dockerized the app and deployed it. App was then used in user tests.· Used ReactJS, JointJS library, Flask, Neo4j, Cypher and Heroku.

Bachelor'S Degree, Computer Engineering

Computer Science

Activities and Societies: Twente Hacking SquadErasmus+ student at the University of Twente