AI Agent for Live Encar Browser Automation – POC

We are a South Korean automotive company developing an AI-powered global car-buying platform. For the first phase, we want to build a small Proof of Concept focused specifically on live browser interaction with Encar.com, a major Korean used-car marketplace. The AI agent should: Understand a customer's vehicle requirements in natural language. Convert those requirements into structured search criteria. Open and browse Encar.com through real browser automation. Apply relevant search filters. Read current search results. Open individual vehicle listings. Extract important information such as model, year, mileage, price, options and available accident/vehicle-condition information. Select several suitable vehicles. Return structured data that an LLM can analyze and explain to the customer. Example: A customer asks: “Find me a BMW 520i, 2022 or newer, under 60,000 km and under KRW 45 million.” The system should understand the request, browse Encar live, find matching vehicles, inspect several listings and return a reliable comparison. This is NOT a simple chatbot or a basic scraping project. We are specifically interested in reliable browser automation integrated with an AI agent. Preferred experience: Python Selenium and/or Playwright Browser automation AI agents / LLM tool calling OpenAI / Claude APIs Backend development Structured data extraction Error handling and production-grade automation If the POC is successful, this can develop into a much larger project involving a complete AI car-buying assistant, international market research, payments, vehicle inspections, shipping, customs information and internal company workflows. Please explain briefly how you would architect the POC and provide examples of similar browser automation or AI-agent systems you have built.