Bwana Mkubwa

I’m a storyteller and dialogue specialist who designs realistic, multi-turn conversations to probe and improve large language models. I blend creative writing, role-playing, and critical analysis to reveal instruction-following gaps, emotional missteps, and coherence failures.

My work centers on adversarial yet plausible prompt engineering and methodical documentation: I craft scenarios, escalate tension naturally, and record failure modes with clear, rubric-based observations. I’ve run self-directed experiments iterating prompts and personas to refine LLM outputs.

I also bring practical annotation and image-labeling experience using tools such as Label Studio, CVAT, Labelbox, and GT Studio, having handled thousands of smartphone and drone photos for agritech models. My academic background supports structured thinking across social and behavioral dimensions.

I’m fully remote-ready, organized for part-time engagement, and passionate about making AI safer and more human-aligned through realistic testing and careful reporting.