can demir

Language model post-training (SFT, preference optimization, RL), small language models, agentic systems, evaluation, credit risk and decision models.