An engineer-run studio
RackHoldings builds its own software and consults on the infrastructure underneath it. Two disciplines, one small team, no account managers in between.
RackHoldings is a small studio working at the intersection of two disciplines most teams keep separate: production AI systems and enterprise network infrastructure.
The engineering background is network and security — Cisco campus and data-centre builds, security architecture, firewall policy consolidation. The kind of work where a single misconfiguration can take down a trading floor. That taught us to design for failure, document everything, and validate by evidence rather than assumption.
More recently that has extended into production AI infrastructure: GPU clusters, model serving platforms, RAG pipelines, and the automation that makes both AI and network infrastructure maintainable. We run this stack daily rather than demonstrating it. Real data, real workloads, real constraints.
The same team builds our own apps — RackHoldings designs, ships and releases native Mac and iOS software, which is where the habit of keeping systems boring gets tested hardest.
Most network engineers can't touch ML infrastructure. Most ML engineers can't design a leaf-spine fabric or debug a vPC peer-link failure at 2am. Doing both is the whole point.
- Background
- 12+ years in enterprise network and security engineering across financial services, healthcare, retail, manufacturing and technology.
- Specialisation
- Network architecture, security assessment, compliance automation, and AI/ML infrastructure design and deployment.
- In production
- DGX Spark and GPU cluster, vLLM/LiteLLM model serving, PostgreSQL with pgvector for RAG, n8n workflow orchestration, Prometheus and Grafana.
- Based
- Decatur, Alabama — remote nationwide.
How we approach the work
PrinciplesThese guide every engagement, from first discovery through to operational handoff.
Documentation as deliverable
Every engagement produces HLD/LLD documents, runbooks and configuration artifacts your team can reference, modify and maintain. If it isn't documented, it didn't happen.
Repeatability over heroics
We build processes and tooling that work the same way every time. Repeatable procedure reduces risk, enables delegation, and makes outcomes predictable.
Minimal blast radius
Changes are scoped to limit impact. Migration runbooks carry rollback triggers, and validation runs before, during and after every change window.
Verify by evidence
Outcomes are validated against design intent with automated checks rather than assumptions. Compliance is demonstrated through live configuration data, not self-attestation.
Build what we preach
We run a production AI platform and ship our own apps daily — GPU serving, RAG pipelines, automated workflows, observability. We don't consult on things we haven't built and operated ourselves.
Want to know if we're a fit?
Email contact@rackholdings.com with what you're working on. If we're not the right people for it, we'll say so.