Case Study: ConnectWave automates product price comparisons with Upstage's purpose-trained LLM

ConnectWave - Customer Case Study

ConnectWave, a leading e-commerce platform in Korea, faced significant inefficiencies with its team manually extracting product information for its price-comparison service. This process was slow and prone to error, and general-purpose large language models could not handle the task with the required domain expertise or security. To address this, ConnectWave partnered with Upstage to develop a specialized solution.

Upstage built a private, purpose-trained LLM for ConnectWave, fine-tuned specifically for attribute-value extraction. Utilizing a Mixture-of-Experts architecture and trained on both public and private data, the model automated data extraction with high accuracy, outperforming models like GPT-3.5. The solution dramatically reduced manual workloads for ConnectWave, improved the standardization of product metadata, and enhanced the overall shopping experience for its users.

ConnectWave