{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Database School","title":"The database for all your AI needs","html":"<iframe width=\"100%\" height=\"180\" frameborder=\"no\" scrolling=\"no\" seamless src=\"https://share.transistor.fm/e/e91e5030\"></iframe>","width":"100%","height":180,"duration":3607,"description":"Marcel Kornacker, the creator of Apache Impala and co-creator of Apache Parquet, joins me to talk about his latest project: Pixeltable, a multimodal AI database that combines structured and unstructured data with rich, Python-native workflows.\nFrom ingestion to vector search, transcription to snapshots, Pixeltable eliminates painful data plumbing for modern AI teams.\n\nFollow MarcelPixeltable: https://pixeltable.comPixeltable GitHub: https://github.com/pixeltable/pixeltableLinkedIn: https://www.linkedin.com/in/marcelkornacker\n\nFollow AaronTwitter: https://twitter.com/aarondfrancisLinkedIn: https://www.linkedin.com/in/aarondfrancisWebsite: https://aaronfrancis.com – find articles, podcasts, courses, and moreDatabase School: https://databaseschool.com\n\nChapters0:00 – Introduction0:20 – Meet Marcel Kornacker1:19 – Early career and grad school in databases2:12 – Joining Google and building F13:42 – How F1 used Spanner at Google4:01 – Starting Apache Impala at Cloudera6:02 – Why SQL still matters7:29 – What keeps Marcel fascinated with databases9:37 – The “SQL is dead” waves and shift to AI10:21 – Observing pain points in computer vision pipelines13:02 – Multimodal data challenges and the idea for Pixeltable16:10 – How Pixeltable handles transformations with computed columns26:29 – Example: processing video, audio, and transcripts in Pixeltable33:12 – DAG execution and parallelism explained37:00 – Transactional guarantees in Pixeltable39:00 – Iterators and chunking data for search42:26 – Using embeddings and semantic search47:05 – Updating data and incremental recomputation50:06 – Thoughts on RAG and hybrid search53:14 – Real-world use cases and dataset curation57:00 – Example: labeling food waste on cruise ships1:02:00 – Labeling workflows and syncing annotations1:02:41 – Pixeltable’s roadmap and cloud vision1:07:10 – How to get involved with Pixeltable1:09:03 – Closing and where to find Marcel","thumbnail_url":"https://img.transistorcdn.com/QPpKhRBWtM56k1sC7Y7j7aqGy8sd-XKWw6mEKgA1jmw/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS9lMmRl/ZTM5ZTAxOTMzNTgy/MmJmZTA4MWE1ZjMw/YjE4Ny5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}