The Full Stack: Scavenger Mode Multi-Agent

People come to me when everyone else has said no.

I don't have a fixed service menu. I don't specialize in one layer of the stack. I work across all of them: from rewriting BIOS firmware on salvaged motherboards, through kernel-level optimization, to building multi-agent AI systems that coordinate autonomous operations. If it runs code, I can make it do what you need.

This is what a single person built, from nothing, on hardware nobody else wanted:

Bare metal: Salvaged DDR3-era hardware. Huananzhi X79 dual-socket with a fully rewritten BIOS. 80GB VRAM across a Vega 64 + 4x Radeon Instinct MI25 array.

System level: Custom Linux kernel builds, PCI lane reconfiguration, ROCm compute stack optimization.

AI inference: Quantized LLMs up to 122B parameters running locally. Mixture of Experts routing. Models that aren't supposed to run on this hardware, running on this hardware.

Multi-Agent orchestration: Hierarchical agent systems with a custom 80B commander model coordinating fleets of smaller specialized agents. Human-in-the-loop approval for all active operations.

Cloud infrastructure: Google Cloud AI Infrastructure specialization in progress (AI Hypercomputer, Cloud GPUs, Cloud NPUs). IBM AI scholarship recipient.

Google Cloud AI Infrastructure specialization (AI Hypercomputer, Cloud GPUs, Cloud NPUs) — in progress

IBM Artificial Intelligence learning path — scholarship recipient

The cloud knowledge isn't a pivot away from bare metal. It's the missing bridge: understanding both sides means I can build hybrid infrastructure that most engineers can't even conceptualize.

When people ask what I offer, I don't have a clean answer. I solve the problems that others won't touch. The ones that cross too many layers, require too much depth, or simply don't fit into anyone's job description.

If you pay me fairly, I'll make it work.