At Isik University, I’m developing FakeShield, an end-to-end pipeline for real-time fake-news detection. I fine-tuned BERTweet for rumour and fake-news classification on the PHEME and LIAR datasets, with semi-supervised continual learning through pseudo-labeling.
I built a reproducible DVC pipeline and tracked experiments and models in MLflow. I also implemented drift monitoring and GitHub Actions workflows for cloud-based retraining, and designed model serving around a FastAPI REST endpoint.

