BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//sf26-agenda//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VEVENT
UID:SESS-217@sf26-agenda
DTSTAMP:20261020T151709Z
DTSTART:20261104T200000Z
DTEND:20261104T202000Z
SUMMARY:Your AI Agents Can't Read Your Iceberg Table: Delete Formats\, Quer
 y Engines\, and Iceberg v4
LOCATION:Innovation Theater
DESCRIPTION:Shubham  Baldava\n\nAI agents now out-query humans 29:1\, and t
 heir median query scans 7K rows\, not 200K. That workload wants cheap embe
 dded engines like DuckDB and PyIceberg reading Iceberg directly. But most 
 Kafka and CDC pipelines write updates as equality deletes\, a format PyIce
 berg can't read\, Snowflake and Databricks reject\, and Iceberg v4 just pr
 ohibited (spec vote passed Aug 2026). Your streaming table is "interoperab
 le" only for Spark. This talk covers how we made an open-source CDC-to-Ice
 berg writer emit positional deletes and v3 deletion vectors at streaming s
 peed. Iceberg has no primary-key index\, so we built one: a per-table LSM 
 (Pebble) key-to-row-location index\, checkpointed against the table's snap
 shot\, that survives Spark compaction and Trino writes to the same table\,
  and migrates legacy equality deletes in one atomic commit. You'll leave w
 ith a 2026 engine-by-delete-format matrix\, the write/read/compaction cost
  of each format\, and a decision framework: query engines\, then format ve
 rsion\, then delete mode. By attending this session\, your contact informa
 tion may be shared with the sponsor for relevant follow up for this event 
 only.
URL:https://sf26.rmoff.net/s/your-ai-agents-cant-read-your-iceberg-table-de
 lete-formats-217
END:VEVENT
END:VCALENDAR