I have three source documents ready for you—the official project guideline PDF, an ICD-10 general-guidelines PDF, and a CPT codes Excel file. Once a user manually uploads an outpatient medical-record PDF, the agent must extract the clinical narrative, map it against those guidelines, and return the recommended ICD-10 and CPT codes. For every code it surfaces, it should also present a concise explanation that cites the relevant rule or note drawn from the supplied references. The flow I picture is straightforward: drag-and-drop upload → automated OCR/NLP parsing → guideline-driven reasoning → code suggestions with rationale → downloadable report (CSV or JSON would be great). If you prefer Python with libraries such as PyPDF2, pdfplumber, spaCy, or a transformer model through Hugging Face, that works for me; just keep the solution self-contained so I can host it locally or on a small cloud instance later. Deliverables • Fully working AI agent (script, model, or small web app) • Source code with clear comments and a brief setup guide • Sample output using one record I’ll provide to confirm ICD/CPT accuracy and explanation quality Acceptance criteria • Correctly identifies ≥90 % of target diagnoses and procedures on my test set of outpatient charts • Each suggested code is paired with a guideline-based justification pulled from the supplied PDFs/Excel • Handles PDFs of varying layouts without manual pre-formatting Let me know your preferred stack and any clarifications you need, and we can get started right away.