Joseph — Agentic Engineer
Move pointer · Select Plan / Act / Verify
An agent is only interesting when its loop can be inspected, corrected, and trusted.
Joseph works on the practical layer of AI systems: tools, memory, handoffs, verification, and the product surfaces that let humans stay in command. The work is less about making a model sound impressive and more about making an automated process legible.
This homepage treats that engineering stance like an editorial object: precise typography, collage imagery, and a restrained Rococo signal of gears, ornaments, and decision paths. The result is personal, but not ornamental for its own sake.
Selected systems
Featured production systems and open-source intelligence architectures developed by Joseph.
AI Agent Runner & Lifecycle Governor
spec-dev
A zero-dependency execution engine and lifecycle governor for Codex and Claude Code, driving agent delivery loops with hard gates.
AI-Powered Relationship & Communication Coach
Love Master
A full-stack multi-modal relationship assistant built with Java (Spring AI, Spring Boot) and React. Integrates screenshot OCR analysis, PgVector RAG knowledge retrieval, and MCP tools.
AI-Assisted Personal Knowledge System
Knowledge Vault
A structured personal wiki repository featuring raw source ingestion, taxonomy organization, and automated LLM compilation via PowerShell workflows.
Availability
Build the loop before the lore
For agentic product systems, evaluation design, workflow automation, and AI interfaces that need to move from prototype to practice.