I design, build, and operate local AI infrastructure, personal intelligence systems, GPU inference environments, and practical automation on user-owned hardware.
My work spans physical infrastructure, multi-node compute, power and networking, model deployment, Windows desktop engineering, release systems, and technical documentation. I take projects from initial planning and hardware assembly through implementation, validation, documentation, and ongoing operations.
I build and operate high-performance local AI environments using NVIDIA DGX Spark systems, RTX 6000-class workstation hardware, and supporting network, power, storage, and automation infrastructure.
This includes hands-on responsibility for:
- Hardware selection, physical installation, and system commissioning
- Multi-node architecture, networking, and cluster validation
- Rack planning, electrical load distribution, cooling, and environmental monitoring
- Local model deployment, quantization, benchmarking, and inference workflows
- Failure detection, guarded recovery, and operational automation
- Security boundaries, reproducibility, evidence capture, and operator documentation
My public engineering lab for local AI infrastructure, multi-node inference, GPU experiments, agent automation, and hardware-tested operator workflows.
An evidence-led recovery architecture for an eight-node DGX Spark environment.
I designed and built the physical power layout, controller mapping, network health checks, safety gates, recovery policy, commissioning process, and sanitized public engineering record. The system distinguishes isolated hardware failures from shared network outages before permitting a guarded power cycle.
Reproducible NVFP4 quantization, evaluation, performance, and local inference workflows for NVIDIA GPU hardware.
A synchronized reference fork of NVIDIA's public DGX Spark playbooks. My independent implementations, measured results, infrastructure decisions, and hardware evidence remain clearly separated in Gumbii AI Lab.
I maintain Windows 11 community editions of three open-source filmmaking applications originally created by Sam Wasserman / Wasserman Productions.
I led the Windows engineering, native packaging, runtime integration, release validation, security controls, documentation, and upstream contribution work across:
The work included Electron and NSIS packaging, Windows filesystem and process behavior, FFmpeg and AI runtime integration, MCP bridges, CI/CD, exact-installer validation, SBOMs, checksums, and software provenance.
I also authored the complete operator documentation library:
- Blockout Operator Handbook
- Motion Previs Studio Operator Handbook
- Stem Studio Operator Handbook
- Three-Application Windows Workflow Handbook
- Thirty-six documented and validated user scenarios
- Implementation, verification, security, troubleshooting, and release records
The reusable Windows contributions were submitted upstream, with the core Windows ports accepted by the original projects.
- Local AI infrastructure and user-owned inference
- NVIDIA DGX Spark and professional GPU environments
- Multi-node compute, networking, power, and recovery systems
- Model deployment, quantization, benchmarking, and validation
- Windows desktop portability and Electron packaging
- Technical writing, operator handbooks, and engineering workbooks
- CI/CD, native QA, SBOMs, provenance, and reproducible releases
- Security, privacy, licensing, and attribution boundaries
Gumbii Digital LLC
North Carolina, United States
X / Twitter
For project-specific questions, open an issue in the relevant repository.