marwan User

Annotated and labeled multimodal datasets for LLM training pipelines, applying guidelines and rubrics to assess correctness and quality. Conducted RLHF preference-ranking and safety-labeling reviews, scoring model outputs against accuracy, instruction-following, and safety criteria.

Selected through Mercor for a text annotation role evaluating LLM outputs for accuracy, safety, and instruction-following on technical, coding, and reasoning tasks. Provided structured written feedback identifying inconsistencies, missing context, and unclear labeling or acceptance criteria in model responses.

Labeled and reviewed recordings and scripts against project annotation guidelines, identifying errors, inconsistencies, and gaps and documenting them clearly for the review team. Collaborated asynchronously with a distributed team across time zones, adapting to evolving annotation standards.

Confirmed high-severity finding in the Olas audit and a medium-severity finding in the Rujira audit, ranking Top 24/96 in the VII-Finance-Contracts audit on Cantina. Discovered and responsibly disclosed a critical information-disclosure vulnerability in a university web system, applying OWASP Top 10 methodology.

Led the social media function for Enactus, planning and reviewing content output and keeping messaging consistent across the team. Coordinated a team of volunteers across multiple student activities over two years, dividing work, reviewing output quality, and meeting event deadlines.

Produced Arabic and English content for client campaigns, writing and adapting copy to brand guidelines and tone-of-voice rules. Supported digital marketing activity across social platforms, including content planning, scheduling, and performance review.

Pursuing a Bachelor of Science in Computer Science, currently in the third year. Coursework includes data structures, algorithms, object-oriented programming, computer networks, and operating systems.