Common challenges include model hallucinations, confusion between similar data types (such as order numbers and tracking numbers), and prohibitively high token costs when processing HTML-formatted email.

In this post, you’ll learn how fine-tuning Amazon Nova models using Amazon SageMaker AI addresses these specific issues by teaching the models to recognize your exact data patterns, distinguish between similar fields, and process information more efficiently—achieving up to 94.77% extraction accuracy while reducing costs 50%.

Parcel Perform worked with the AWS Generative AI Innovation Center (GenAIIC), which provides business and technical consultancy throughout the customer journey. Working backward from Parcel Perform’s problem statement, the team scoped a project to optimize Nova models through various customization techniques and parameter optimization.

This collaboration allowed concurrent improvement of multiple metrics: accuracy, latency, and cost. Le Vy, AI Team Lead at Parcel Perform, reported that the fine-tuned Nova Micro models achieved up to 94.77% extraction accuracy on the testing dataset, an improvement of up to 16.6 percentage points over the baseline. The fine-tuned Nova Micro reduced inference latency by more than 30 percent and halved costs compared with Parcel Perform’s previous model, while matching or exceeding the fine-tuned Nova Lite model at lower cost. With these results, Parcel Perform moved the solution into production to improve its e-commerce logistics operations.