Full-stack dev with a Java/JS background, now deep in local-first AI tooling — LLM routers, agentic coding loops, on-device inference. I like taking a stack apart until it runs on my own hardware instead of someone else's API bill.
- 🔭 Currently building LocalRouter — a local-first LLM router with intelligent failover and error-aware provider blocking
- 🤖 Deep in agentic coding: Claude Code, Ralph-style autonomous loops, Obsidian vaults driven by AI agents
- 🎲 Side quest: AI-assisted tabletop RPG tooling — NexusCampaigns, pathway2e
- 🌱 Learning ROCm/HIP for AMD GPU inference — see llama.cpp-moe-amd-lowvram
- 👨🏻💻 Most of my projects are on GitHub
- 📫 Reach me on LinkedIn
37 public repos. Top languages by repo count: Java, JavaScript, Python, TypeScript, Go, Rust, PowerShell, C#, Vue, CSS.
| Repo | Description | Language |
|---|---|---|
| LocalRouter | Local-first LLM router with intelligent failover and error-aware provider blocking |  |
| easyforkjoin | Easier implementation of forkJoin for RxJS |  |
| NexusCampaigns | AI-powered nexus that connects ideas into complete, interconnected tabletop RPG campaigns |  |
| llama.cpp-moe-amd-lowvram | Run MoE LLMs (e.g. Qwen 3.6 35B-A3B) on 24GB AMD GPUs via llama.cpp + ROCm/HIP |  |
| ProjectCell | Brazilian initiative to learn biology through a game |  |
| CopyPaste2Diffusion | Clipboard image straight to any Stable Diffusion API — paste and go |  |