Amira is an AI reading tutor, serving five million students across four thousand districts in all fifty states, and independently validated to deliver impact comparable to high-quality human tutoring. Building on our existing partnership, CHOP will lead a large-scale clinical study collecting multimodal reading data from 300+ clinically diagnosed children — capturing speech, timing, and engagement signals across structured at-home sessions with autism, ADHD, and speech-impairment populations alongside neurotypical controls. UPenn’s Linguistic Data Consortium will transform that data into the first research-grade, openly accessible corpus of atypical child oral reading — annotated for disfluencies, atypical prosody, attention patterns, and engagement markers. We then use that corpus to train new adaptive algorithms directly inside Amira, improving speech recognition for atypical voices, building attention-aware pacing for students with ADHD, and delivering personalized multimodal scaffolds that meet neurodiverse learners where they are.