At Zipline, I label and review imagery and sensor data, check map data for accuracy and consistency, and validate outputs in real time. I document recurring issues and edge cases, flag anomalies, and work with ML engineers to align annotation guidelines with system needs.
I’ve also evaluated machine learning models through projects at CMU Africa and Hugging Face. For KinyarMedASR, I developed a Kinyarwanda medical speech recognition model and reduced word error rate by 7% through preprocessing and spell correction; I also documented failure cases for Qwen3.5-0.8B-Base.

