Deep Dives

Technical articles on distributed systems, schema governance, and ML pipelines. Each article is rooted in real engineering experience — referencing actual code, Apicurio Registry contributions, and production architecture decisions.

Apicurio Registry as an Apache Iceberg REST Catalog

Apicurio Registry 3.2.0 now implements the Apache Iceberg REST Catalog API, turning your schema registry into a table catalog for Spark, Trino, DuckDB, and more. One service for both schema governance and lakehouse metadata.

Iceberg Apicurio Data Lakehouse Spark Trino

Real-Time vs. Batch ML Inference: Engineering Tradeoffs You Won't Find in Tutorials

The inference spectrum isn't binary. Micro-batch, dual-model architectures, PySpark UDF tracing challenges, and what changes at 100x scale — lessons from building a real-time classification pipeline.

ML Inference Spark Kafka Distributed Systems Observability

Model Agreement as a Proxy for Ground Truth in Streaming ML

When you deploy ML models in a streaming pipeline, you don't have labels. Dual-model agreement rate, disagreement taxonomy, confidence calibration, and drift detection without ground truth.

ML Model Monitoring Streaming Prometheus Distributed Systems

MCP Servers for Domain-Specific AI Tooling: Lessons from Apicurio Registry

Building an MCP server that wraps a schema registry for AI agent use. Designing the tool surface, security considerations, and patterns for wrapping any REST API as an MCP server.

MCP AI Agents Apicurio Schema Registry LLM Tooling

Designing Schema Evolution for Event-Driven Systems

Schema evolution is the hardest coordination problem in event-driven architectures. Compatibility modes, common mistakes, and real-world patterns from building a Kafka-based ML pipeline with Apicurio Registry.

Schema Evolution Kafka Apicurio Event-Driven Architecture

Building a RAG Chatbot with Schema Registry as the Knowledge Backend

Using Apicurio Registry to store versioned prompt templates as registry artifacts, then feeding them to a RAG chatbot. A unique intersection of schema governance and LLM tooling that nobody else has explored.

RAG LangChain4j Gemini Apicurio LLM

Why Schema References Break in Avro (And How to Design Around It)

Avro schema references promise reusable types across schemas. In practice, they break in subtle ways — circular dependencies, version pinning drift, namespace collisions. Here's what I learned contributing to Apicurio Registry.

Avro Schema Registry Apicurio Distributed Systems