At Ethara AI, I evaluate and annotate AI/LLM outputs against structured criteria for instruction following, factuality, relevance, and response quality. I review model behavior across large volumes of content and document errors and inconsistencies.
I apply labeling and evaluation guidelines to support accuracy and consistency in post-training workflows. My work includes prompting, tokens, model behavior, and RLHF/RLVR concepts.
Through independent OSINT research, I practice verifying information across public web sources and documenting evidence. I also study web and API security using OWASP methodologies, TryHackMe learning rooms, and authorized labs.
