Skip to content

// field notes

What I learned shipping the thing

No 'AI news roundup' spam. I write when I've actually built something worth writing about — usually with the diagrams I made debugging it.

AllAI (4)Voice AI (1)Automations (1)
How to Build a Production-Ready Multi-LLM System: A 2026 Architecture Guide coverAI
AI 8 min read

How to Build a Production-Ready Multi-LLM System: A 2026 Architecture Guide

A deep architecture guide to multi-LLM systems — model routing, fallbacks, cost instrumentation, and caching — from someone who runs these in production and cut a client's model bill 40–60%.

Building an AI Voice Calling Agent: A Complete 2026 Walkthrough coverVoice AI
Voice AI 7 min read

Building an AI Voice Calling Agent: A Complete 2026 Walkthrough

How to build an AI voice calling agent that holds real phone conversations — the STT to LLM to TTS pipeline, sub-second latency, interruption handling, and clean human handoff. Built from a live production system.

RAG Explained: Building Retrieval-Augmented Generation with LangChain coverAI
AI 6 min read

RAG Explained: Building Retrieval-Augmented Generation with LangChain

A practical LangChain RAG tutorial that goes past the demo — chunking strategy, embedding choice, hybrid search, evaluation, and the source-citation grounding that keeps a chatbot from making things up.

n8n vs Make: Which Automation Platform Should You Use for AI Workflows in 2026? coverAutomations
Automations 6 min read

n8n vs Make: Which Automation Platform Should You Use for AI Workflows in 2026?

A hands-on n8n vs Make comparison for AI automation — pricing, AI nodes, self-hosting, error handling, and which one I actually reach for on client builds, with a clear decision rule.

FastAPI for AI Apps: Serving LLMs in Production Without the 2am Pages coverAI
AI 7 min read

FastAPI for AI Apps: Serving LLMs in Production Without the 2am Pages

How to serve LLMs in production with FastAPI — async streaming endpoints, auth, rate limiting, caching, and observability. The production scaffolding I rebuilt one too many times, explained.

Choosing a Vector Database in 2026: pgvector vs Pinecone vs Chroma coverAI
AI 7 min read

Choosing a Vector Database in 2026: pgvector vs Pinecone vs Chroma

A practical vector database comparison for RAG — pgvector vs Pinecone vs Chroma on cost, scale, ops, and filtering. Which one I default to, when I switch, and the decision rule I use on client builds.

Get the build notes

New posts and what I learned shipping them, 1–2× a month. No spam.