Blog and insights¶
This is where I publish implementation notes, architecture lessons, and delivery heuristics from building AI systems that need to work outside the demo environment.
The focus is practical: how to make agentic workflows more dependable, how to improve retrieval quality, and how to ship AI APIs that internal teams can operate.
What you will find here¶
-
Agentic workflow design
Patterns for deterministic orchestration, tool usage boundaries, human review, and failure containment.
-
RAG and hybrid retrieval
Practical retrieval decisions covering ranking, fusion, evaluation, and why many RAG systems underperform in production.
-
Production AI delivery
API contracts, FastAPI service design, deployment boundaries, observability, and what changes once real users arrive.
Featured posts¶
-
AI-Powered Email Automation: From Chaos to Action
The public webinar version of a production email automation system built with PydanticAI, RAG, and FastAPI.
-
Production playbook for agentic AI workflows
A practical checklist for deciding what should stay deterministic, what can be delegated to the model, and how to keep the system testable.
-
Why hybrid retrieval wins more often than pure vector search and how to structure the system so ranking can evolve safely.
-
Service boundaries, streaming contracts, background processing, and the operational details that matter once traffic is real.
Want these ideas applied to a real system instead of a blog thread?¶
If your team is already experimenting with AI but needs stronger architecture decisions, that is exactly the gap I help close.