@netflix Our session is on Day 2, Nov 5 at 4pm. Check out the agenda here:

@netflix We’ll walk through how Ray enables large-scale processing across hundreds of GPUs, while LanceDB’s columnar design provides efficient, intelligent curation and sampling. Together, they’re producing smaller, more diverse, and higher-quality datasets for cutting-edge text-to-image

@netflix In this session, we’ll share how Netflix is transforming multimodal data curation using Ray for distributed ingestion, filtering, and inference across massive video and image corpora. All powered by LanceDB as the high-performance storage and query layer.

Building and curating large-scale multimodal datasets has long been a complex, resource-heavy challenge. But that’s changing fast. Lei Xu of LanceDB and Pablo Delgado of @netflix will be speaking at Ray Summit 2025 — Scaling Multimodal Data Curation with Ray and LanceDB

I’ve been a huge fan of @pacoid for over a decade. Really excited to see the mention of @lancedb for entity resolution and linking in critical investigative applications (also, @tech_optimist !)

@netflix 5/5 The Multimodal Lakehouse lets developers focus on data and experimentation, not infrastructure. No more brittle DAGs or duplicated pipelines — just unified multimodal data at scale. For a full deep dive into the Multimodal Lakehouse, read the blog 👉

@netflix 4/5 It enables teams to: • Store + query structured & unstructured data side by side • Run feature engineering + training on multimodal datasets • Scale distributed compute using Python + Ray • Use hybrid search (vector + full-text + SQL) across captions, frames, and

@netflix 3/5 The LanceDB Multimodal Lakehouse is an AI-native architecture built to handle text, images, video, audio, and structured metadata all in one system. It merges the scalability of a lakehouse with the low-latency performance of a vector database.

@netflix 2/5 Why the Multimodal Lakehouse? Traditional data lakes were built for structured metrics tables. They break down when you add multimodal workloads: embeddings, videos, audio, captions, annotations.

1/5 @netflix leverages the Multimodal Lakehouse to power its Media Data Lake — unifying video, audio, subtitles, and metadata into one foundation for workloads like scene understanding, feature extraction, localization, and semantic search.

Welcome! Excited for what we will be doing together!

All proceeds from PyData Seattle support @NumFOCUS, the nonprofit behind Pandas, NumPy, and Jupyter. Come learn, connect, and build with the open-source data community in Seattle!

Supercharging Multimodal Feature Engineering with Lance and Ray by Jack Ye, Senior Software Engineer 📅 Sat, Nov 8 | 11:40 AM Learn how Lance + Ray power scalable multimodal AI pipelines with efficient storage, distributed compute, and fast iteration.

Data Loading for Data Engineers by Weston Pace, Senior Software Engineer 📅 Fri, Nov 7 | 10:55 AM Discover how smarter data loading impacts training speed and model quality—and how LanceDB boosts ML workloads with caching and random access. 🚀

@changhiskhan, CEO & Co-founder of LanceDB, will deliver a keynote at #PyDataSeattle! Chang helped build Pandas and is now leading the next generation of data infrastructure for AI. Don’t miss it. 💡 📅 Nov 7–9 | Bellevue College

🚀 Excited to announce that LanceDB is headed to @PyData Seattle 2025! Join us Nov 7–9 at Bellevue College for 3 days of AI, machine learning, and open-source innovation with 400+ data scientists, engineers, and builders. 🌟 🎟️

Instead of prompts to generate art, discover what human's have already created. Find art to match a feeling, a poem, or a dream… Explore with your imagination at https://t.co/wVN9y2YaRP

art should be made by humans, not robots. https://t.co/VEFtcuEAu8 is a search engine for the masterworks of human creativity. free for everyone, forever.

Excited to share a passion project we've been working on that rethinks how we discover art. In an age of AI-generated images, we wanted to use AI to build a bridge back to human-made art. The result is @semanticdotart, a multimodal art discovery engine that lets you search

7/7 Many thanks Many thanks to @pavitarsaini, @lychkel, @MischaLamoureux, @adam__conway, @panth_bhavsar for their ideas and work on this project!