Writing
Notes from production AI
What I'm seeing in real engagements: RAG patterns, agent orchestration, evaluation discipline, and the architecture decisions that emerge when AI ships into regulated environments.
The hybrid sovereignty path nobody talks about (and why it usually beats both extremes)
Most teams default to strict self-hosting or full pragmatic frontier APIs when EU sovereignty enters the room. The third path — a hybrid routing layer between open-weight bulk operations and EU-region frontier APIs for hard reasoning — usually wins for regulated mid-market. Here's the architecture.
How I Use Claude Code + MCP to 10x My Engineering Workflow
I use Claude Code with MCP servers daily for AI development, task management, and life automation. Here's my exact setup and what actually works.
Why I Migrated from LangChain to LangGraph (and What I Learned)
I migrated a RAG copilot serving 5,000+ users from LangChain to LangGraph. Real code, real metrics — the langchain vs langgraph comparison you needed.