Software Architect · Bend, OR
Brian Love
Agentic software architect building developer tooling for fullstack AI-powered web applications.
I build the systems and tooling behind AI-native product work.
Most of my work lives where product strategy, frontend architecture, developer experience, and AI interaction design overlap. I care about software that is useful in production, understandable to the team, and honest about the tradeoffs.
Writing and speaking are extensions of the same craft for me. They help me test ideas in public through articles, conference talks, and other media so the useful patterns can travel further than one team or one product cycle.
Faith, family, and place shape the kind of work I want to do.
I am a Christian, and that shapes how I think about ambition, responsibility, and the kind of work worth doing. I want technical excellence and personal integrity to point in the same direction.
I am married to Bonnie and we are raising our daughter in Bend. Life here has room for church, family, friends, and long days outside skiing, biking, hiking, and resetting in the mountains.
Writing
Software architecture, AI engineering, and the craft of building tools.
Featured
ai
Agentic Memory and What It Means for Web Apps
Agentic memory changes web apps from thin shells around stateless prompts into systems that can learn across sessions, adapt behavior, and manage context as product infrastructure.
ai
Building an Azure AI RAG Pipeline with SharePoint
A practical guide to building a production RAG pipeline that indexes SharePoint documents into Azure AI Search, chunks intelligently, generates embeddings, and grounds LLM responses in real organizational data.
ai
Google A2UI: Fixed Schemas, Dynamic Schemas, and a Safe Fallback Strategy
A practical architecture for building fullstack agent web apps with Google A2UI v0.9 using fixed catalogs by default, dynamic schema overlays for the long tail, and deterministic fallback when validation fails.
ai
The Frontend Reward Loop for Agentic Software
Agents improve through interaction. The frontend is where intent, correction, and outcomes are visible together, and where prompt augmentation can take most teams surprisingly far before offline reward-model training.