General
Beelze: A Tiered Safety Policy for a Desk AI Agent
The safety design of an AI assistant built to reach mail, calendar, and a computer: risk tiers that fail closed, a staged confirmation the model…
Case study category
General
The safety design of an AI assistant built to reach mail, calendar, and a computer: risk tiers that fail closed, a staged confirmation the model…
Nov 10, 2026
General
The workflow I use to run AI coding agents across my projects: EARS requirements, a handoff document as a hard gate, a reviewing model that…
Nov 10, 2026
General
The AI infrastructure behind my own engineering work: one local endpoint for several model providers, an orchestrator with a human gate and hard caps, and…
Nov 10, 2026
General
A local CrewAI pipeline that reads real data, writes and executes Python, and builds HTML dashboards, validated by whether the code runs rather than by…
Feb 18, 2026
General
A career intelligence platform that joins O*NET occupational data, SEC EDGAR workforce investment signals, and BLS wage statistics, with cosine-similarity skill matching and an ATS…
Feb 18, 2026
General
A CrewAI pipeline of specialized agents, coordinated by a state machine, turns CSV data into validated Python code and HTML dashboards, with schema checks at…
Feb 18, 2026
General
A RAG pipeline that turns a large archive of workforce education documents into a semantic knowledge base, using structure-aware chunking, hybrid keyword and embedding retrieval,…
Feb 18, 2026