Frankline Mutuma

AI Training Specialist and LLM Evaluation Professional with 4+ years of experience supporting LLM training, Generative AI evaluation, prompt engineering, RLHF, SFT, DPO, human preference ranking, model alignment, AI safety, red teaming, hallucination detection, bias analysis, and AI quality assurance. Experienced in creating, evaluating, and validating high-quality training data across text, image, video, audio, and speech. Skilled in rubric-based evaluation, instruction-following assessment, factuality and contextual accuracy review, multimodal annotation, ASR labeling, error classification, and structured human feedback for continuous model improvement.

AI Model Evaluation Specialist at TELUS International AI (2025-01 – 2026-01)

• Evaluated LLM responses for accuracy, relevance, safety, consistency, instruction following, and contextual quality.
• Performed RLHF pairwise ranking, prompt evaluation, hallucination detection, bias analysis, and rubric-based scoring.
• Provided structured human feedback and reviewed challenging prompts and edge cases to support model improvement.

AI Trainer & Computer Vision Annotator at Outlier AI (2022-01 – 2024-12)

• Created and reviewed AI training datasets and evaluated prompts for LLM and Generative AI workflows.
• Annotated and validated text, images, video, and audio, including image segmentation, object tracking, and ASR labeling.
• Maintained 98%+ annotation quality while meeting high-volume project standards.

AI Data Collection & Validation Specialist at Freelance (2022-01 – Present)

• Curated, cleaned, validated, and quality-checked AI training datasets.
• Identified annotation and model-quality issues, developed error taxonomies, and produced validation reports.
• Applied detailed evaluation guidelines consistently across AI data projects.

Master of Arts in Sociology (Criminology and Social Order) – University of Nairobi (2019)

Bachelor of Science in Information Technology – Jomo Kenyatta University of Agriculture and Technology (JKUAT) (2015)