Osman Eray Erdoğan

At Marmara University, I developed a cross-domain PSD estimation pipeline that compared five estimators across QPSK-OFDM, EEG, and audio noise. The findings showed that LMS adaptive pre-filtering does not uniformly improve PSD accuracy and form the basis of a paper in preparation. I automated the pipeline’s results in Python, generating Excel and CSV reports alongside heatmaps and spectrograms. I also designed a multi-metric framework to compare the methods. As an AI Research Intern at Huawei Technologies Co. Ltd., I prepared training datasets and trained large language models using five training methods. I evaluated outputs and internal representations with Explainable AI methods and documented findings for ongoing research. At CRC Elektronik, I verified DC power system units through functional and load testing and contributed to production validation. I currently draw and plan transformer-substation electrical schematics at Armtek Elektrik using EPLAN Electric P8.