Infrastructure & systems engineer. I run the entire technical stack for a 25-person bioinformatics company — from the datacenter rack to the AWS bill to the marketing CRM. One person, production systems, no safety net.
Nine-plus years from service desk to sole-charge infrastructure.
I'm the sole IT staff for a small bioinformatics software company, which means everything is mine: the on-premise datacenter rack, the AWS environment, identity and MFA, workstations, networking, vendor management, and the backup strategy that lets me sleep at night.
My career arc runs Tier I support → systems analyst (including two years in Frankfurt as part of a first-generation Follow-the-Sun support model) → systems administrator in a PCI/HIPAA-regulated fintech environment → sole-charge infrastructure. Along the way I picked up a bias for automation: if I do it twice, it gets an Ansible role or an OpenTofu module.
In 2026 I also took ownership of the company's technical marketing operations — CRM administration, the website, email campaign infrastructure, and Google Ads / technical SEO. This site is part of that portfolio: built, deployed, and instrumented by me.
Production outcomes, not lab exercises. Employer details available on request.
Planned and executed migration of company storage onto an enterprise NAS appliance — staged sync, verified integrity, and cut over live systems without losing a byte or a workday.
Brought AWS EC2 and network configuration under OpenTofu across environments, and standardized Linux host configuration and centralized authentication with Ansible.
Designed and drove MFA requirements across all user accounts — selection, rollout sequencing, exception handling, and the human side of getting 25 people through an auth change.
Coordinated an office and datacenter connectivity migration across a 3-month window — circuit provisioning, failover planning, cutover scheduling, and vendor wrangling.
Racked, cabled, and integrated a new hypervisor node into the production cluster — hardware through hypervisor config to workload rebalancing.
Took over the technical side of marketing operations after a staff departure: CRM administration and data flows, website operations and deployments, email campaign infrastructure (deliverability, segmentation, automation), and Google Ads with technical SEO — conversion tracking, Search Console, structured data.
Spent two years in Europe designing and launching a PCI/HIPAA-regulated fintech's first Follow-the-Sun support model — building the operational framework and absorbing the escalations that would otherwise have paged US on-call overnight.
Coordinated an unplanned shutdown of the fintech's Frankfurt datacenter on short notice for building maintenance — managing the controlled turndown and the full recovery afterward.
Built internal training and cross-team collaboration protocols across development, networking, database, infosec, and support — turning repeat production escalations into documented, self-served procedures.
Working demonstrations of how I think agentic tooling and modern infrastructure should be built — with scope limits, approval gates, and guardrails treated as first-class features. Write-ups and repos land here as each one ships.
Expose a NetBox instance to Claude via MCP so an agent can answer "what's racked where, on which VLAN, owned by whom" from documented truth — read-only by design, with the schema doing the guardrail work.
Stack: NetBox, MCP server, Claude, home-lab inventory.
An agent that reads an inbound queue, classifies, drafts responses, and proposes actions — but every state-changing step requires human approval. The PoC is as much about the gate design as the automation.
Stack: Claude, MCP, ticket system API, audit log.
Codify real operational runbooks (service restart, disk pressure, cert renewal) as agent-executable procedures with explicit boundaries: read-only diagnostics free, mutations gated, destructive actions blocked outright.
Stack: Claude Code, Ansible, systemd, allowlist policy layer.
An agent that reviews OpenTofu plan output before apply — flags destructive changes, drift, and cost implications in plain English. Human stays on the apply button; the agent makes the plan readable.
Stack: OpenTofu, Claude, CI pipeline.
A destroy-and-rebuild-from-scratch lab: hypervisor, VLANs, DNS, identity, and services — all defined in OpenTofu + Ansible. The repo is the documentation; the rebuild is the proof.
Stack: Proxmox or similar, OpenTofu, Ansible, Git.
Metrics, logs, and alerting over the reference stack — dashboards that answer real on-call questions, plus alert rules tuned to signal rather than noise.
Stack: Prometheus, Grafana, Loki, Alertmanager.
jamessingley.com built as a demonstration: hand-rolled single-file site on Cloudflare Workers static assets, JSON-LD structured data, Search Console instrumentation, and Core Web Vitals as the benchmark. The write-up covers each decision.
Stack: Cloudflare Workers, HTML/CSS, schema.org, Search Console.
SPF, DKIM, and DMARC configured from scratch on this domain, with a monitoring loop on DMARC reports — the unglamorous plumbing that decides whether campaigns reach inboxes.
Stack: Cloudflare Email Routing, DMARC reporting, DNS.
The working stack, grouped by what it's for.