Inspiration

Modern B2B procurement often suffers from opaque pricing, unverified vendor metrics, and manual checkout friction. We built AutoProcure AI to bridge this gap by bringing automated, intelligent deal evaluation and instant payment workflows together into a single unified platform.

What it does

  • Scikit-Learn ML Pipeline: Automatically analyzes product categories, warranty tiers, seller ratings, and review counts to predict fair market value, calculate a 0-100 deal score, and deliver an instant recommendation verdict (e.g., Strong Buy / Undervalued).
  • PayPal Sandbox Checkout: Seamlessly triggers automated B2B test orders and secure simulation flows to complete the procurement loop instantly.

How we built it

  • Backend & ML Engine: Python, Scikit-Learn, Pandas, FastAPI, Uvicorn, Joblib, and Python-Dotenv.
  • Frontend UI: HTML5, Tailwind CSS, and asynchronous JavaScript (Fetch API).
  • Payment Layer: PayPal Sandbox API integration supporting both live sandbox tokens and safe fallback simulation mode.

Challenges we ran into

Handling absolute path resolution for the serialized Scikit-learn model (deal_scorer_pipeline.joblib) inside FastAPI, synchronizing asynchronous frontend payload calls with backend endpoints, and ensuring robust fallback mechanisms for the PayPal checkout loop.

What we learned

Building a production-minded machine learning pipeline from raw data generation to real-time web deployment, and integrating enterprise payment sandbox APIs cleanly.

Built With

Share this project:

Updates

Submission history