PayPilot-AI

Your Autonomous AI Shopping and Payment Assistant

Comment

Inspiration

Online shopping often requires users to visit multiple websites, compare products manually, evaluate prices, and make purchasing decisions across different platforms.

I wanted to simplify this experience by building an AI-powered shopping assistant that goes beyond product recommendations. The idea behind PayPilot AI was to create an autonomous commerce agent that can understand shopping requirements, research products, compare options, evaluate budgets, and help users complete purchases securely.

What it does

PayPilot AI is an autonomous AI-powered commerce agent that transforms natural-language shopping requests into an intelligent purchasing workflow.

Its key capabilities include:

  • Natural Language Product Discovery: Users describe what they need, and the system discovers relevant products.
  • Intelligent Product Comparison: Compares products based on pricing, specifications, and relevance.
  • AI-Powered Recommendations: Recommends suitable products based on user requirements and budget.
  • Multi-Agent Orchestration: Coordinates specialized AI agents for discovery, comparison, decision-making, security, payments, and monitoring.
  • Security Validation: Validates purchase details and requires explicit user approval before payment.
  • PayPal Integration: Uses PayPal Orders API v2 for order creation, buyer approval, and payment capture.
  • Transaction Ledger: Records completed transactions and provides transaction history with CSV export.

How we built it

PayPilot AI was built using a React and Vite frontend with a Python FastAPI backend and PostgreSQL database.

The AI architecture is led by NVIDIA Nemotron 3 Ultra 550B A55B, supported by specialized models:

  • NVIDIA Nemotron – Lead reasoning and orchestration.
  • Mistral Ministral 14B – Product comparison and analysis.
  • Groq-hosted GPT-OSS 120B – Fast supporting inference.

Channel3 is used for product discovery.

The payment workflow integrates PayPal Orders API v2. The backend creates orders, handles buyer approval, captures payments, verifies capture results, and records transactions.

The architecture follows a human-approved purchasing model: the AI assists with decisions, but the user explicitly approves the purchase before payment capture.

Challenges we ran into

One of the biggest challenges was coordinating multiple AI agents while maintaining a consistent purchasing workflow.

Other challenges included:

  • Normalizing product prices across different currencies.
  • Keeping product discovery and budget validation consistent.
  • Implementing secure payment approval and backend-side capture verification.
  • Handling PayPal Sandbox regional payment restrictions.
  • Ensuring transaction records remain consistent after payment capture.
  • Separating AI recommendations from actual payment authorization.

These challenges helped shape a more reliable architecture with explicit approval gates and backend-controlled payment processing.

Accomplishments that we're proud of

The biggest accomplishment was successfully connecting the complete AI shopping workflow with PayPal payment processing.

We successfully demonstrated:

  • Natural-language product discovery.
  • AI-powered product comparison and recommendations.
  • Multi-agent orchestration.
  • Budget and security validation.
  • Explicit buyer approval.
  • Successful PayPal Sandbox payment capture of $26.52 USD.
  • Verified payment confirmation.
  • Transaction persistence in the Transaction Ledger.
  • CSV transaction export.

Most importantly, we demonstrated the complete journey from product discovery to successful payment capture.

What we learned

Building PayPilot AI taught us that autonomous commerce requires more than intelligent recommendations.

A reliable commerce agent must combine AI reasoning with security controls, accurate pricing, explicit user consent, payment verification, and persistent transaction records.

We also learned how to integrate multiple AI models into a coordinated architecture and connect AI-driven decisions with real payment workflows using PayPal Orders API.

What's next for PayPilot AI

Our next goal is to make PayPilot AI more capable and scalable by introducing:

  • More advanced personalized product recommendations.
  • Improved product discovery across additional merchants.
  • Stronger spending controls and purchase preferences.
  • Enhanced transaction monitoring and reconciliation.
  • More sophisticated multi-agent decision-making.
  • Broader commerce integrations and production-ready payment infrastructure.

Our vision is to make online shopping simpler, smarter, and more secure through autonomous AI-powered commerce.

Built With

Share this project:

Updates

Submission history