BioAdapt: An open-source multimodal infrastructure platform for autism-related physiology research

BioAdapt is an open-source, local-first Python and FastAPI research-infrastructure platform for multimodal autism-related physiology research. The platform normalizes heterogeneous wearable, EEG, event-related-potential, environmental, and consumer-export data into a shared canonical schema; exposes quality-control and preprocessing assumptions; supports exploratory profile reporting; and includes executable claim-language and raw-data-boundary safeguards. This v1 preprint describes the current BioAdapt architecture, parser-readiness evidence, exploratory baseline results, and repository safeguards. Verification of the current development branch collected 305 tests, all of which passed. The BCIAUT-P300 path reaches metadata and BIDS-sidecar parser readiness across 15 subjects and 210 records. The Zenodo infant EEG PSD path parses feature-level exports for 281 participants and yields weak exploratory macro-F1 values that do not support individual-level inference. Move4AS reaches metadata and local-structure parser readiness. A targeted PhysioNet Wearable Exam Stress replay preserved quality-control warnings and was not promoted to a clean acceptance pass. BioAdapt is research infrastructure only. It does not support diagnosis, infant screening, clinical decision support, behavior prediction, hospital deployment, or regulatory claims. Stronger claims would require longitudinal autism-specific validation, independent replication, prospective governance, and community review.