More than ten years across high-performance computing, distributed systems, and applied artificial intelligence. I build the platform layer that real users depend on — not the demonstration. Currently shipping two products on my own hardware: Foundry, a bare-metal hardware control plane, and Agent MA, an artificial intelligence agent workforce platform.
I build artificial intelligence platforms across the full stack — from the hardware fleet at the bottom up through the agent layer at the top. The list below is what I am hands-on with today.
I design and ship artificial intelligence platforms in production — agent lifecycle management, multi-provider routing, real-time delivery, observability, and the security plumbing real users depend on. The platform layer, not the demonstration.
I designed and built multi-agent orchestration with agent-to-agent messaging, shared context, and self-healing infrastructure. Pioneered the Model Context Protocol approach in production before it was an established pattern.
I have operated high-performance-computing infrastructure at scale — more than 700 compute nodes, more than 400 artificial intelligence accelerators, three datacenters, more than 800 engineers served. Strong on the real distributed-systems tradeoffs: latency, scale, reliability, cost.
I built bare-metal hardware automation from scratch — provisioning, deployment, and configuration pipelines using open-source tooling. I run production Kubernetes on my own hardware fleet today.
I built retrieval-augmented-generation systems that deflected 70 percent of operational-support ticket volume and gave a non-technical investor-relations team 95 percent answer accuracy on complex queries. I take artificial intelligence capabilities from interesting paper to shipped code that moves the business.
More than twenty years of hands-on hardware experience — motherboard and printed-circuit-board design at AMD, failure analysis at IBM, Base Management Controller firmware at ClearCube Technology, and board bring-up and validation at HP TippingPoint. My platform work rests on hardware fundamentals I built from the ground up.
I have led platform-engineering teams through the transformation from reactive information-technology operations into modern platform engineering. I mentor underperforming engineers into senior contributors. Cross-functional collaboration across hardware, software, machine-learning training, firmware, and business teams.
I have built two products entirely by myself — Foundry (the bare-metal hardware control plane) and Agent MA (the artificial intelligence agent workforce platform). Both run live in production on my own hardware fleet today.
Both products are live and running on my own hardware fleet. Foundry handles the servers themselves. Agent MA runs the artificial-intelligence agents on top. Each works on its own; together they take an empty rack of computers all the way to a working artificial intelligence workforce ready to do real work.
A clean, opinionated control system for the physical servers in a data center. Foundry sits on top of the open-source Metal-as-a-Service provisioning tool and adds what is actually needed for day-to-day operation: a real dashboard, machine-type templates, bulk deployments across many servers at once, a browser-based terminal for any machine, safe snapshots and rollback for upgrades, and a programmable interface that other software (or artificial-intelligence agents) can drive.
Foundry manages network configuration, storage layouts (including high-reliability redundant setups), virtualization (running isolated workloads on shared physical hardware), production-grade clustered Kubernetes (the industry standard for running modern applications), automated post-deployment configuration, role-based access control, and live updates on the state of every server. Today: 52 features shipped, 262 documented programming-interface endpoints, and the ability to perform bulk operations across up to 20 servers in a single command.
Where it fits in the stack: between the low-level open-source primitive that speaks to the bare hardware below and Agent MA (the artificial-intelligence layer) above. One key, one audit trail, one source of truth for hardware operations.
Every user gets their own persistent personal artificial-intelligence assistant. That assistant can create specialist sub-agents on demand — a software developer, a designer, a project manager, a research agent — whichever the user needs for the task at hand. The agents communicate live with the user and with each other, in real time. The platform recovers automatically from any failure with no human operator needed.
Users connect their own artificial-intelligence subscription (their own Claude account, OpenAI account, and so on), so the user pays their artificial-intelligence provider directly — no markup added by the platform. Enterprise-grade security is built in from day one: each user's data is fully isolated, all secrets live in a dedicated secrets-management system, and every action is recorded in an audit trail.
More than 30 artificial-intelligence agents currently running live in production. More than 20 features shipped in the last three months. The platform itself is developed daily by the same artificial-intelligence agent team it sells — the live, observable proof that the model works. Available at agent.gqma.org.
Where it fits in the stack: the top layer of the stack, built last on top of Foundry. The agents reason about what to do; Foundry handles the hardware operations below.
I have managed teams, operated more than 700 compute nodes across high-performance-computing and artificial-intelligence training clusters, and built artificial intelligence agents in production before it was an industry category.
Designing and building Agent MA — a multi-agent artificial-intelligence platform. The platform runs on its own production infrastructure (containerized applications orchestrated by Kubernetes, dedicated secret management, full observability). More than 20 features shipped in the first three months. The platform is built by the same artificial-intelligence agent team it is designed to support.
Owned the strategy and roadmap for the company's high-performance computing and artificial-intelligence training-cluster infrastructure, serving more than 800 engineers across three globally distributed datacenters. Ran second through fourth tier operational support and 24/7 incident response across more than 700 compute nodes and more than 400 artificial-intelligence accelerators. Introduced bare-metal hardware automation from scratch and led the transition from a reactive operations model to a proactive platform-engineering model. Pioneered artificial-intelligence agents in production that reduced operational-support volume by 70 percent — before it was an industry category.
Built near-real-time data-analytics pipelines for global factory operations. Architected continuous-integration-and-delivery systems spanning multiple cloud providers and isolated, air-gapped customer environments. Led zero-data-loss platform migrations. Infrastructure work at scale across every layer, from the database up to the deployment system.
Earlier-career work spanning software and infrastructure architecture, board bring-up and validation on network-security appliances, base-management-controller firmware on blade servers, structured failure-analysis on enterprise systems, and motherboard design and processor verification on early silicon.
Available for consulting on artificial intelligence platforms, high-performance computing, and bare-metal infrastructure — and open to the right full-time role. Or, if you just want to see what an artificial-intelligence agent team running in production looks like, try Agent MA directly.