PhD in signal processing; working on AI and machine learning since 2014, in academia and industry. My current focus is the reliability of AI evaluation: auditing AI judges, graders and leaderboards so that reported numbers can be trusted. Open-source work at https://github.com/vahit19: reward-audit, rl-env-audit, judge-ceiling, leaderboard-audit, rater-audit and runopsy. Earlier work: GenAI and agentic systems (LLMs, VLMs, RAG), GPU image and video generation pipelines delivered as production APIs, and AI-based sports video analytics as a remote contractor for SportsMEDIA Technology (US). Based in Istanbul. LinkedIn: https://www.linkedin.com/in/vahit-feryad-19517256/