Predictive Export Clearance System

Замовник: AI | Опубліковано: 23.10.2025
Бюджет: 30 $

Our export-authorization desk still operates on manual checks, and it’s hurting us. Each snack variant, destination market and MOQ deviation triggers a maze of compliance rules; every delay means higher inventory holding and occasional demurrage fees. I need a production-ready predictive system that will tell the team—before we pick and pack—whether a sales order can ship, and under which export code or compliance note. Scope of work • Build or fine-tune an ML model that predicts “Authorize / Reject / Flag for Review” for every order line. • Ingest historical shipment data, variant attributes, destination-specific regulations and MOQ patterns; cleanse and engineer features that capture ingredient, packaging and labeling constraints. • Wrap the model in an API (Python FastAPI, Flask or similar) so our SAP middleware can call it in real time. • Dashboard the key metrics—probability score, governing rule hits, and recommended corrective actions—so planners can override when necessary. • Optimise for three success targets we will track in UAT: – Cut authorization turnaround time by 60 %. – Reach a minimum 298 % accuracy uplift versus our current rule sheet. – Help trim inventory-holding and demurrage costs by 15 %. Data & environment CSV extracts from our WMS/SAP (≈120 k rows, three years), a compliance rule matrix in Excel, and region-specific regulation text. You can spin up on AWS or work locally; we already license Snowflake and S3. Preferred stack: Python, scikit-learn / XGBoost / LightGBM, plus Docker for portability—but I’m open to stronger alternatives if it helps meet the KPIs. Deliverables 1. Clean, documented codebase with reproducible training pipeline. 2. Containerised inference service with REST endpoint and basic auth. 3. Streamlit dashboard with real-time model output and rule highlights. 4. Deployment guide and hand-off walkthrough. Acceptance will be based on a side-by-side pilot against live orders during a two-week window. If you’ve tackled export, customs or supply-chain compliance models before, your insight will be a major plus.