20+ years of building stuff — from Linux supercomputers & Top 500 HPC clusters → Web 2.0 startups → award-winning AI platforms for the US Navy.
- 🔭 Now: AI Factory Lead at a Fortune 500 company
- 🎖️ Before: Lead AI Infrastructure Engineer at Helix Center for Applied AI — built an award-winning AI platform for the US Navy (Jan 2025)
- 🧠 Way back: Cycorp Inc. — reported to Dr. Doug Lenat, worked on symbolic AI for defense & healthcare
- 🚀 Early days: Sold Linux supercomputers, built Top 500 HPC systems, infrastructure at LiveJournal & OpenDNS
- 🏆 NVIDIA Certified: DGX SuperPOD · AI Infrastructure · InfiniBand
- 🌆 Community: Founder, Detroit AI Collective
| ⭐ | Project | Lang | What It Does |
|---|---|---|---|
| 🥇 | RegicideOS | 🦀 Rust | AI-native Linux distribution (Gentoo + Btrfs + Cosmic Desktop) |
| 🥈 | zerochain | 🦀 Rust | Multi-agent orchestration via plain text files |
| 🥉 | merlin | 🦀 Rust | RL-based LLM prompt router (sub-ms async routing) |
| 🔧 | opencode-pushguard | 📘 TS | Pre-push git hook for AI-assisted workflows |
| 🛡️ | opencode-watchtower | 🐚 Shell | Safer agent tool calling with utcp isolation |
| 🧠 | opencode-memento | 📘 TS | Preserves project context across AI coding sessions |
| 🔮 | Xfiles | 🌐 Protocol | Modern Plan 9 protocol with quantum AI reinforcement learning system |
Full list over at PROJECTS.md.
The repos above are just the surface. The full stack spans autonomous agents, distributed systems, LLM infrastructure, OS research, and confidential computing:
| Project | Domain | What It Does |
|---|---|---|
| Nexus | 🤖 Autonomous Agents | Multi-agent orchestration with self-improving capabilities and recursive task decomposition |
| Routage | 🌐 Distributed Systems | Intelligent message routing, mesh networking, and fault-tolerant cluster coordination |
| Merlin | 🧠 LLM Infrastructure | End-to-end pipeline for model training, fine-tuning, quantization, and high-throughput serving |
| RegicideOS | 💻 Operating Systems | Experimental kernel with AI-native syscalls, scheduler awareness for inference workloads, and capability-based security |
| DCAP | 🔒 Confidential Computing | Remote attestation frameworks, secure enclaves, and zero-trust verification for AI workloads |
| AAS-Monad | 📐 Formal Methods | Mathematically-grounded scoring framework for self-improving systems |
This work is anchored in the belief that the future of intelligence is unified, not fragmented. It explores ideas around AI Pangaea—the convergence of distributed intelligence islands into a more cohesive, secure, and capable continuum.
- Superintelligent Entities — Implemented agents with advanced cognitive architectures, persistent memory, and recursive self-improvement
- AI Security at Every Layer — Tool-calling sandboxing, LLM input/output guardrails, multi-agent consensus verification, and confidential execution
- Memory & Continuity — Long-horizon agent memory systems that persist across sessions, devices, and intelligence islands
Read the paper: AI Pangaea: Unifying Intelligence Islands — one of many inspirations for me recently.
This repo (awdemos/demos) is where I dump production-grade patterns I actually use:
- Kubernetes — 100+ deployment patterns, Cilium eBPF, multi-cluster
- LLM Infrastructure — RAG pipelines, fine-tuning, multimodal benchmarks
- CI/CD — Dagger Go pipelines with container-native builds
- IaC — Pulumi (Go/TypeScript) for Azure & AWS multi-tenant infra
- Rust — Matrix operations, CLI boilerplate, async patterns
git clone https://github.com/awdemos/demos.gitReproducible. Observable. Composable.
I like building from first principles. I design systems the way Unix intended: small, focused tools that do one thing well and compose cleanly. I'm into symbolic AI and systems thinking — not just chasing the latest hype. I don't trust big tech consolidation much, and I'm a big believer in AI sovereignty: models you own, data you control, infrastructure you can actually audit.
I almost always reach for a general-purpose programming language when solving problems — no YAML-only solutions if I can help it. I pull in everything I've learned from a decade of DevOps, decades of systems automation, and decades with Linux to keep things reproducible and maintainable.
My formative years were spent on Slashdot and deep in open source culture — that ethos stuck. I've been contributing to open source for decades.
Memory safety and zero-cost abstractions aren't just Rust features — they're how I think about infrastructure.
© 2026 Andrew White