“One Pipeline to Rule Them All: Unifying Multimodal and AI Data Processing” - 🎥 Our AI_dev Europe talk recording is now live! Last month, @Sammy Sidhu and @Colin Ho demonstrated how to process 1,000 PDFs in 2 minutes at AI_dev Europe, the same time it typically takes to process just one. The problem: Modern AI workloads require processing images, documents, and embeddings at scale, but existing tools force teams to juggle Spark, custom scripts, and multiple frameworks. Teams waste months building infrastructure instead of solving actual problems. In this talk from Amsterdam, they show how Daft eliminates this complexity: • Live demo processing 1,000 PDFs across 768 CPUs and 64 GPUs • Zero OOM errors, zero configuration headaches • Same code runs on your laptop or petabyte-scale clusters • Native support for documents, images, embeddings—any modality Watch the full talk to see how unified multimodal processing actually works in production: