Angel Conde·Sep 25Accelerating Spark with the Java Vector APIApache Spark has had a columnar execution direction for some time, and projects such as Apache DataFusion Comet have demonstrated how much…
Angel Conde·Aug 28Compiling DuckDB’s AWS extension to WebAssembly: querying Amazon S3 Tables from a browser tabDuckDB-Wasm can’t load the aws extension today. Three small patches later, a React app in my browser was committing Iceberg tables to…
Angel Conde·Aug 25Running DBT over Spark Connect on Glue 6.0 with Iceberg v3 andAWS Glue 6.0 interactive sessions now speak Spark Connect. That single change turns a managed Spark cluster into something a local process…
Angel Conde·Jul 20The Case of the Dying Executors: Debugging a Native Memory Leak in Apache Iceberg v3How a suspected deletion-vector problem turned out to be a one-line-of-concept leak in a brand-new code path and how an AI coding agent (…
Angel Conde·Jul 8Running dbt on Amazon EMR Serverless with Spark Connect and S3 TablesFor years, the pitch for dbt was simple: write SQL, get a warehouse-native transformation framework. The catch was always the warehouse. If…
Angel Conde·Jun 23Spark RTM mode — Low latency streaming pipelinesFor its entire life, Spark Structured Streaming has been a micro-batch engine. It feels continuous, but under the hood it slices your…
Angel Conde·Jun 9Accelerating Apache Spark with Apache DataFusion CometApache Spark is widely adopted for large-scale data processing, but many of its performance optimizations have plateaued. Spark’s…
Angel Conde·Sep 17, 2025Iceberg on DuckDB: End-to-End Example with Amazon S3 TablesApache Iceberg has quickly become the de-facto standard for building LakeHouses on AWS. It brings reliability and ACID guarantees to object…
Angel Conde·Sep 12, 2025Accelerating Apache Spark with Gluten & VeloxApache Spark is widely adopted for large‑scale data processing, but many of its performance optimizations have plateaued. Spark’s…A response icon1A response icon1
Angel Conde·Mar 26, 2025Using Debezium and Kafka Connect with Iceberg ( part II)On part I we have seen how we can deploy an end to end solution for CDC using Open Source libraries that can be deployed easily using…A response icon1A response icon1
