Sangwa Michel

I’m an AI/ML engineer and data scientist focused on multilingual NLP, LLM benchmarking, and RLHF-style evaluation. My work spans Kinyarwanda, English, Kiswahili, and French, with a strong emphasis on adversarial testing and qualitative error analysis—especially where evaluation data is scarce.

At the CMU Open Learning Initiative (RISE Project), I apply machine learning and data analysis to education-focused research in a cross-functional setting. I’m also pursuing an MS in AI and Data Science at Carnegie Mellon University Africa, backed by strong competition results on real-world datasets.

My projects reflect end-to-end benchmarking and stress-testing: I built an automated benchmark for Kinyarwanda–English QA, designed adversarial prompts to probe safety and robustness in African-language LLMs, and developed an ASR pipeline for low-resource languages. I enjoy turning messy, multilingual signals into clear evaluation evidence that teams can act on.