Over the past couple of years I ve been building infrastructure for data platforms and AI systems. My background is in analytics engineering, cloud data platforms, PostgreSQL, Snowflake, and AWS, but lately my focus has shifted toward building products rather than consulting.
My current project is pg-cdc, an open-source PostgreSQL Change Data Capture platform that continuously publishes changes to modern data lake formats like Parquet and Apache Iceberg. The motivation came from seeing teams spend weeks wiring together Debezium, Kafka, connectors, and custom pipelines when many simply want reliable, governed data flowing into their analytics or AI stack.
More broadly, I m fascinated by how AI agents will consume operational data. I believe the next generation of software won t just answer questions it will perform work autonomously. That requires fresh, trustworthy, permission-aware business context, and I think data infrastructure is becoming a critical part of the AI stack.
We ve found that the hardest part of building AI applications isn t the model it s getting governed, trustworthy operational data into AI systems. How are you handling change data capture today? Are you streaming directly from PostgreSQL, building custom pipelines, or relying on periodic ETL?
pg-cdc is an open-source PostgreSQL Change Data Capture engine that streams WAL changes directly to Apache Iceberg and Parquet on Amazon S3. Unlike traditional CDC pipelines, it runs as a single binary—no Kafka, Spark, or JVM required. The output is immediately queryable with Athena, DuckDB, Spark, Trino, and other Iceberg-compatible engines, making operational PostgreSQL data ready for analytics and AI workloads.
