Founder & AI Systems Developer (RLHF & Conversation Evaluation)
As Founder & AI Systems Developer at Refyul, contributed to reinforcement learning from human feedback and related labeling workflows. Designed and implemented RLHF-style prompt evaluation and conversation correction processes to improve AI agent performance. Created, tested, and refined AI-powered conversational agents, focusing on data labeling for prompt/reply accuracy and behavior evaluation. • Built and operated an AI-powered WhatsApp sales agent using multi-step conversation review. • Developed internal evaluation loops for continuous agent improvement in real-world scenarios. • Labeled and corrected model responses based on structured conversation logic and state management. • Authored comprehensive documentation to support data annotation and evaluation compliance.