If blogs aren't your thing, we also put our Flotilla story into a short video: • Why multimodal data pipelines break Spark and Ray Data • How Daft makes it easy to build pipelines for PDFs, images, audio, and video • The architecture behind Flotilla: Swordfish workers on every node, coordinated at cluster scale • Real benchmarks across audio transcription, document embedding, image classification, and video object detection — where Daft is 2-8x faster than Ray Data and 4-18x faster than Spark • What Flotilla unlocks for terabyte-scale workloads Watch the full video here: