Writing
Agentic systems in production.
Technical writing on system design, retrieval architecture, orchestration patterns, and the decisions that separate production AI from prototypes.
They built it with AI first. Then they restarted all of it.
Why FluidCloud rebuilt its multi-cloud migration engine without AI — deleting four months of LLM-based mapping for deterministic rules. A field guide to where LLMs belong in infrastructure.
The signal is the easy part. Here's what's hard.
How Signalbase runs a production signal pipeline: waterfall verification, confidence gates, feedback loops, and where human judgment still bottlenecks GTM execution.
Why your RAG pipeline works in staging and fails your users
You tuned retrieval for two weeks. The demo impressed the team. You shipped it. Three weeks later, users were getting answers that had nothing to do with their questions.
The orchestration layer that doesn't collapse under its own weight
It starts clean: a planner, three tools, a simple loop. Six months later you need four files open to explain what it does. This pattern helps avoid that.
Memory in agentic systems: four layers most teams collapse into one
Your agent forgets what it should remember. The fix is not a bigger context window. You need four memory layers, each doing a different job.