Architecting multi-agent engineering workflows, local-first AI runtimes, and cloud systems on Azure.
Focused on transforming emerging agentic capabilities into production-ready developer platforms, enterprise FinOps automation, and collaborative multi-agent harnesses.
| Project | Description | Role & Metrics | Tech Stack |
|---|---|---|---|
| bond | Multi-agent harness CLI to coordinate pi, Claude Code, Codex, Copilot, opencode, Crush, and Gemini using Herdr and Git worktrees. | Author & Maintainer Documentation |
TypeScript · Node.js · Herdr |
| Azure/CCOInsights | Continuous Cloud Optimization Insights: enterprise dashboard and automated framework for FinOps and governance in Microsoft Azure. | Maintainer |
Azure REST API · Power BI · FinOps |
| squadcenter | Desktop mission control application for orchestrating GitHub Copilot CLI sessions across multiple Squad agent teams with live telemetry. | Author & Maintainer Code & Demo |
Electron · React 19 · TypeScript |
| fluzo | Local-first agentic programming runtime with fine-grained permissions, budget boundaries, and explicit cancellation mechanisms. | Author (Active R&D) Project Overview |
Rust · TUI · Local Agents |
Selected analyses and field notes from josecorral.dev:
- bond: Multi-agent coordination in action — Architectural design behind orchestrating multiple coding agents across isolated git worktrees.
- Applying RPI strategies to AI development agents — Building common skills and execution gates for reproducible agentic coding.
- Renewing the local AI stack for agents — Benchmarking local LLM inference, llama-swap, Docker, and Halogen for agent runtimes.
- Vulkan vs ROCm on AMD Strix Halo — Evaluation of llama.cpp inference backends on unified memory architectures.





