Portrait of Dr. Sara Owusu
ML evaluation lead

Who is Dr. Sara Owusu? ML evaluation lead at Food Tracker Compass

Machine learning evaluation lead

PhD Machine Learning, University of Toronto / Vector Institute · Toronto, Canada · she/her

Sara is responsible for everything reproducibility-related at Food Tracker Compass. She owns the double-grading protocol — every score on the site is independently graded twice, with disagreements adjudicated against the reference set — and the statistical sign-off that decides whether a measured difference between two apps is actually large enough to be reported. Before joining the team she spent three years at AWS AI Labs on the Bedrock evaluations group and two years at Cohere on model quality. She is the final technical reviewer on every individual app review published on this site.

What is Dr. Sara Owusu's area of expertise?

Model evaluation methodologyStatistical reproducibilityMultimodal calibrationHuman-rater agreement

Where did Dr. Sara Owusu study?

  • PhD, Machine Learning — University of Toronto (Vector Institute affiliate) (2019)
  • MEng, Electrical & Computer Engineering — University of Cape Town (2014)

Where has Dr. Sara Owusu worked previously?

  • Applied Scientist, AWS AI Labs — Bedrock evaluations team · 2021 – 2024
  • Research Engineer, Cohere — model quality · 2019 – 2021
  • Research Intern, Microsoft Research Cambridge — distributed ML · 2018

What has Dr. Sara Owusu published?

  • Double-Grading Protocols for Multimodal Benchmark Datasets — NeurIPS Datasets & Benchmarks, 2023
  • Calibration of Vision-Language Models Under Distribution Shift — ICML, 2022
  • Inter-Rater Agreement in Food-Composition Estimation — Journal of Food Composition and Analysis, 2021

What food tracking topics does Dr. Sara Owusu research?

Model evaluation methodology, statistical reproducibility, calibration of large multimodal models.

Who else reviews food tracking apps at Food Tracker Compass?