Apache Iceberg™ Meetup Bay Area

Apache Iceberg™ Meetup Bay Area! 🧊❄️ Join the Iceberg community on Monday, November 2nd at Google's Sunnyvale office for an evening of Apache Iceberg talks, co-hosted by Google, VeloDB, and Datastrato. Connect with fellow enthusiasts, share insights, and dive into the latest developments in the Apache Iceberg™ ecosystem. Whether you're a seasoned pro or new to Apache Iceberg, this meetup is the perfect place to exchange ideas and spark innovation. Agenda 5:00 PM - 5:45 PM: Doors open, networking, and food 5:45 PM - 7:15 PM: Welcome remarks and presentations 7:15 PM - 8:00 PM: More networking The event will focus on use cases around and innovations in Apache Iceberg (https://iceberg.apache.org/). Speaker Lineup Code as Data: Turning a Codebase into Business Logic on Iceberg Bin Ni - Co-founder & CEO at DeepMorph AI A codebase isn't really a pile of files - it's a graph. In DeepMorph, we built a graph that captures beyond syntactic facts, but also the layers above them: high-level user scenarios, data journeys, and business logic. We model our enterprise codebase as a graph and store it in Apache Iceberg, serving traversals through an API → Cypher → graph engine → Iceberg path. That upper layer is where it gets interesting beyond engineering. Because the graph links code to the data it touches and the business scenarios it serves, it doubles as a lineage and impact-analysis tool: a data scientist or risk analyst can trace which business processes depend on a given data field, find where a regulatory rule actually lives in the system, or measure the blast radius of a change - all as graph queries rather than weeks of manual archaeology. Two kinds of consumers read that graph. Agents query it through an MCP server to understand the codebase faster than crawling raw files. Engineers, analysts, and data scientists use a product that traverses and summarizes subgraphs to make sense of the system and its changes.

Hive to Iceberg at Roku: a REST catalog rollout with Apache Gravitino Bharathkrishna Guruvayoor Murali - Senior Software Engineer at Roku Most Hive-to-Iceberg migrations stall on the same question: how do you move thousands of tables that hundreds of jobs depend on, without asking every team to change code? This case study shows how Roku is doing it in production.

We split the work into two migrations that can each be rolled back alone. The first moves tables from Hive to Iceberg, which gives us atomic commits, schema evolution and managed partitions. The second moves the metadata layer from Hive Metastore to an Iceberg REST catalog, which gives us per-user identity, scoped short-lived storage credentials and one catalog across clouds. Apache Gravitino sits in front of HMS, and HMS stays behind it as the source of truth while tables move. Every job still writes spark_catalog.db.table, so no SQL or DAG changed.

The talk focuses on what was hard: Keeping both catalogs consistent with metadata event hooks and a sync service, and what we do when they disagree Spark and Trino fallback, so a query resolves correctly whether a table has moved or not Converting table-level Hive grants into RBAC Why Trino's per-user token is unsigned, and the token proxy we built so the Iceberg REST endpoint can verify who is calling You will leave with a staged migration plan you can reuse, the failure modes we hit, and the design decisions we would repeat or change. It is aimed at engineers running Hive, Spark or Trino who are evaluating Iceberg REST catalogs. Engineering leads will get a clear view of the sequencing, the risk and what "done" looks like.

Beyond Tables: Open Semantics with Apache Ossie and Polaris Yufei Gu - Staff Software Engineer at Snowflake Apache Iceberg enables different engines to share the same tables, but a BI dashboard and an AI agent can still answer the same business question differently. “What was our revenue last quarter?” leaves two important choices unresolved: gross or net sales, and calendar or fiscal quarter. As definitions spread across SQL, BI models, and agent instructions, keeping their meaning consistent becomes a shared engineering problem.

This talk follows an illustrative retail example to explore an open approach to semantic interoperability. We will use Apache Ossie (incubating) to express datasets, relationships, metrics, and business context, then examine how Apache Polaris can provide a common catalog home for those models alongside Iceberg tables and views. A concrete YAML example and a walkthrough of Polaris’s beta semantic-model REST API will connect the model format to discovery, retrieval, controlled updates, and access permissions.

We will also examine the boundaries: sharing a model does not guarantee identical query results, and permission to read a model does not grant access to its source data. Consuming tools must preserve calculation semantics and resolve time periods correctly.

Attendees will leave with an architectural framework for sharing business definitions across BI and AI tools, an understanding of current integration constraints, and a view of community directions in converters, ontologies, and semantic governance.

Getting there Building entrance, parking, and check-in details will be posted here closer to the event. Please have a valid, government-issued ID ready at the front desk. About Google Cloud Google Cloud is committed to an open Data Cloud built on open source, open standards, and open formats like Apache Iceberg. BigQuery and BigLake give data teams an AI-ready lakehouse on Iceberg: BigLake tables for Apache Iceberg on Cloud Storage, BigLake Metastore with an Iceberg REST catalog, and Iceberg tables that are as easy to query as native BigQuery tables, whichever platform created them. 📚 Read about Google Cloud's commitment to Apache Iceberg 🖥️ Get started with BigQuery About VeloDB VeloDB is a commercial data warehouse powered by Apache Doris, an open-source real-time analytics database. Founded in May 2023 by original contributors to Apache Doris, VeloDB provides enterprise-grade real-time analytics products and solutions to global customers. The team is composed of 70% R&D personnel and is led by long-time contributors to the Apache Doris project, including engineers with deep expertise in database systems and cloud computing. 📚 Learn more about Apache Doris 🖥️ More at velodb.io About Datastrato Datastrato is the original creator of Apache Gravitino, the open-source metadata lake that federates the catalogs you already run, including Iceberg tables through its Iceberg REST Catalog, under one namespace with shared access control and lineage. Datastrato also hosts the Data for AI meetup series in the Bay Area. 📚 Learn more about Apache Gravitino 💬 Join the Gravitino community on Discord 📲 Follow Datastrato on LinkedIn 🖥️ Explore demos and talks on Datastrato's YouTube channel 🌐 More at datastrato.ai