github.com/datafusion-contrib/StreamFusion Hi all! This project is still in its infancy, but I've been working on building the first OSS flink accelerator! Under the hood, we parse existing flink table API plans and then operate on arrow batches instead of rowdata. I've built new versions of most operators using datafusion, or on top of existing code from arroyo/risingwave to keep the code fast! I'm still iterating here to get even benchmarks but feel free to follow along, contribute, or just leave a star! I plan to keep this OSS and ideally donate it to the datafusion project on a longer term basis.