Pushing local Gemma 4 through a multi-file refactor, rerunning it under opencode, and racing Claude Code. The best security lesson came from the policy log.…
Running a coding agent in YOLO mode, powered by a local Gemma 4, inside a Docker Sandbox microVM. Everything that broke on the way, and the small YAML file that fixed it.…
What changed after running a production RAG system for four months: hybrid search, reranking, and hardening lessons for Java and Kubernetes stacks.…
Why Kubernetes for LLM workloads: GPU scheduling, autoscaling, and serving models like Gemma in a production-grade Java RAG system. Part 3 of the series.…
Architecture deep dive of a production RAG system in Java 25 and Spring Boot WebFlux: service boundaries, retriever design, and tradeoffs explained.…
The real production challenges of RAG systems: latency, reliability, cost, quality, and observability. Part 1 of building production-grade RAG in Java.…
Hands-on intro to DJL (Deep Java Library): building a speech recognition app with an engine-agnostic deep learning framework for Java developers.…
How Pixie brings instant, eBPF-powered observability to Kubernetes: debug services, spot bottlenecks, and profile apps without changing code.…