At xAI, I evaluated LLM outputs for factual accuracy, relevance, and safety, and documented model failure patterns. I also ranked competing responses using RLHF principles to support reward model calibration.
In my data science projects, I built a customer churn classification pipeline that achieved about 82% accuracy and used feature importance analysis to identify churn drivers. I also developed a Prophet-based sales forecasting model and evaluated it with RMSE and MAPE.
I built a Books to Scrape project to extract and clean book pricing and rating data, then explored the results with visualizations. My physics background at Gebze Technical University gave me a foundation in statistics, linear algebra, and scientific research.

