Worked on the Shopify/tidb repository to enhance the efficiency of data ingestion workflows, focusing on the Lightning import process. Refactored the Parquet file sampling logic to sample the average row size once per table rather than per file, which reduced sampling overhead and improved the speed and accuracy of data size estimation for restore operations. This approach enabled calculation of total data size using the sampled average row size and total row count, streamlining restore planning and execution. The work leveraged Go for implementation and drew on skills in data import, file processing, and performance optimization to deliver measurable efficiency improvements.