Worked on the apache/spark repository to enhance PySpark streaming usability by introducing admission control for custom streaming data sources. Focused on data streaming and Python, the work delivered comprehensive documentation and a runnable example demonstrating backpressure-aware ingestion and predictable micro-batch sizing. The implementation included detailed guidance on getDefaultReadLimit() and latestOffset() interactions, helping users understand ReadMaxRows-based batching strategies. Validation was performed end-to-end in Databricks Dogfood Staging, confirming consistent batch sizes and reliable performance across multiple micro-batches. The contribution also improved observability by verifying outputs in the Streaming UI, supporting users in adopting robust admission control techniques in PySpark.