I need an end-to-end AI platform that can ingest clinical notes, discharge summaries, lab results, and any other electronic medical records, then automatically extract the key data points and deliver a concise, clinically useful summary for each patient. The core of the job is building robust natural-language processing and information-extraction pipelines that perform accurately across diverse record formats (PDF, HL7, FHIR JSON, handwritten-OCR where possible). I’ll provide a sample dataset and guidance on which fields matter most; you design the architecture, train or fine-tune the models, and wrap everything in a clean, auditable API or lightweight web dashboard. Python is preferred, and I’m comfortable with libraries such as spaCy, Hugging Face Transformers, PyTorch, or TensorFlow—use what you feel delivers the best balance of speed and accuracy. Because we are handling protected health information, the solution must keep data in encrypted storage and log every access for compliance; no third-party calls that could leak PHI. Deliverables (acceptance criteria): • A working prototype that accepts raw medical records and returns structured JSON plus a readable narrative summary. • Model performance report with precision/recall on the supplied validation set. • Deployment instructions (Docker or similar) so I can spin it up on our internal server. Future phases may extend to pattern recognition or predictive analytics, so modular design is a plus. If you have experience with clinical NLP or HIPAA-compliant DevOps, let me know in your proposal along with an estimated timeline for the MVP.