Developed an end-to-end GLiNER2 demo and Kafka ETL example for the modal-labs/modal-examples repository, focusing on structured extraction and text classification from streaming data. Leveraged Python and Kafka to implement a micro-batch ETL workflow that consumes messages and posts processed results to a REST sink, enabling real-time data processing with structured outputs. Improved code reliability and maintainability by aligning with ruff linting standards and refactoring imports to isolate heavy dependencies, allowing for local execution without unnecessary overhead. Demonstrated collaborative development practices while accelerating the delivery of NLP capability demos and supporting data-driven insights for downstream applications.