Building custom LLMOps: When Langfuse wasn't enough for production AI agents
After evaluating Langfuse, LangSmith, and Phoenix, I built my own observability stack. PostgreSQL + structured logs = observability that works.
Blog
Architecture, AI and technical decisions that matter
MCPs vs custom workflows: Choosing how much freedom your agents get
MCPs give full autonomy. Custom workflows give control. Choosing wrong will cost you months debugging agents that break your business logic.
Hybrid cloud architecture: When self-hosted + cloud is the answer
Why your AI stack shouldn't be 100% dependent on OpenAI/Anthropic nor 100% local. Hybrid architecture as a real solution.
n8n vs Zapier: Why self-hosted wins in AI projects?
Real comparison between n8n self-hosted and Zapier for AI automations. Costs, control and vendor lock-in.
PostgreSQL + pgvector: The underestimated stack for AI
Why PostgreSQL with pgvector can replace your expensive vector database stack? Real cases.
3 fatal errors that kill your AI SaaS before launch
The most common architecture errors I've seen in AI SaaS projects. And how to avoid them from day 1.