I need an AI-driven tool that can turn nothing more than a list of CSV column headers from banking-related financial data into a detailed risk assessment in under eight seconds. Paste the headers, hit “Run”, and the model should return an executive-ready report that flags key financial risks, quantifies their significance, explains the rationale, and suggests follow-up audit procedures. Scope • Input: CSV headers from banking datasets (balance sheet, P&L, GL extracts, etc.). • Processing: Large Language Model (GPT-4, Claude, or comparable open-source alternative) chained with a rules/ontology layer specific to banking audit standards (Basel, COSO, PCAOB). • Output: A comprehensive, well-structured risk analysis—no high-level summaries; I want full detail that an audit partner could sign off on. Key requirements • Sub-8-second end-to-end response time for typical files (≈50–250 columns). • Clear segregation of inherent, control, and detection risk, with scoring logic exposed in the code. • Web interface or lightweight desktop app where I can paste headers or upload the CSV. • Written in Python; using LangChain, FastAPI, Streamlit, or similar is fine as long as dependencies stay mainstream. • All prompt engineering, taxonomy libraries, and custom code handed over with brief but precise documentation and a README that lets me redeploy on my own GPU/CPU box. Deliverables 1. Running prototype hosted on your sandbox (URL or demo video). 2. Source code and environment files. 3. Risk taxonomy/ontology file tailored to banking financial data. 4. Deployment guide and short user manual. Acceptance criteria • I paste a header set from a real bank GL extract and receive a multi-section report (risks, rationale, audit steps) in ≤8 s. • At least 90 % of identified risks align with those in my benchmark answer sheet. • No personally identifiable data leaves the tool during processing. If you have shipped anything similar—AI knowledge engines, audit analytics, or high-speed LLM apps—let’s talk.