I'm obsessed with the surface between humans and machines — terminals, editors, and the AI agents that increasingly live inside them. I build the tools I wish existed, and I like going all the way down: I have built a kernel, a browser engine and an LLM inference engine to understand how they really work.
- Currently building → a multi-provider coding agent, a desktop dev workspace, and a bare-metal microkernel
- Currently learning → Rust internals, GPU runtimes, distributed agent orchestration
- Open to → collaborating on developer tools, AI infra, and systems projects
| Project | What it is | Stack |
|---|---|---|
| Wingman | Terminal coding agent — bring your own model (Anthropic, OpenAI, Gemini, Ollama, OpenRouter) | Rust |
| dlm | Local LLM inference engine that streams transformer layers to run models larger than your VRAM | Rust |
| Arc | Desktop dev workspace: PTY terminal, code editor, git, SSH and a REST client in one window | Tauri · React |
| Echo | Privacy-first voice keyboard — press a hotkey, speak, and it types into any app | Tauri · Rust |
| Kernel Panda | Bare-metal x86_64 microkernel: SMP, capability IPC, NVMe/AHCI, a copy-on-write filesystem | no_std Rust |
| IRA | Voice-first AI assistant with wake word, barge-in, and hybrid local/cloud inference | Rust |
More projects
| Project | What it is | Stack |
|---|---|---|
| zero | AI-native web browser with its own rendering engine | Rust |
| kortex-memory | Multi-tenant memory layer for AI coding agents, served over MCP | Python · pgvector |
| mcp-gateway | One local daemon in front of all your MCP servers, with policy and an audit log | TypeScript |
| Spark | Self-hosted PaaS — push to Git, get a routed app on your own Kubernetes cluster | Rust · K8s |
| mediaserver | Self-hosted media server with on-the-fly transcoding | Go |
| Aegis | Autonomous multi-agent penetration testing in isolated sandboxes | Python |
| clifi | Terminal-native personal finance tracker with a ratatui dashboard | Rust |



