Carvana approached us to help them with optimizing and automation of their entire data collection and management process – from crawling to publishing cars. Our team developed a scalable and stable solution to pull vehicle information from different sources and an optical character recognition service to convert different types of documents such as PDF into readable HTML. Additionally, we built custom services to automatically bid in auctions and enrich data by calling external APIs, parsing description text to detect damage, or calling internally built machine-learning libraries for predicting auction prices or vehicle health status. 

Solution

The system processes around 370,000 entries in each cycle, so we built a custom solution to speed up the process, reducing the time required for a scraping cycle from around 3 hours down to 50 minutes.