The workflow
Structured intake feeds a LightGBM model through a FastAPI backend. Safety rules can escalate the routing suggestion. A React/TypeScript interface supports assessment, review, dashboard and PDF reporting workflows.
A deliberate boundary
A model output is one input into the review process. Rule-based escalation takes priority when defined safety criteria are met; clinician review remains central. This capstone is not a validated clinical deployment.
Engineering beyond prediction
The implementation connects model artifacts, request validation, persistence, review decisions and report generation. That end-to-end path makes error handling and traceability part of the system, rather than leaving the model in a notebook.
What this demonstrates
The project illustrates the founder’s ability to connect Python inference with a usable application and explicit decision boundaries. It is separate from Dreamview client work; no commercial clinical outcome is claimed.