SRMPmachine is a joint effort of the American Museum of Natural History and Massachusetts Institute of Technology designed to teach high school students machine learning (ML) through scientific inquiry. The project innovates within the Science Research Mentoring Program (SRMP) by creating a 175-hour “Summer Institute in ML,” followed by mentored-research experiences using ML. SRMPmachine focuses on ML, a key subset of AI that allows machines to learn patterns from data to make predictions or generate content. The Summer Institute has allowed 120 students to work with ML tools, evaluate models, and address bias by investigating problem sets informed by both natural science and larger societal issues. A subset of 30 students has spent over 130 hours during the academic year working on an ML-enabled research project. Preliminary results suggest that the Summer Institute increased students' knowledge of core ML concepts and interest in AI careers. This increase was sustained throughout the mentored research year among all students, whether or not they used ML in their research. Ongoing interviews are exploring students' attitudes toward AI, self-efficacy, persistence, and sense of belonging.