Platform · AI infrastructure
Jack Devlin
I build the infrastructure AI products run on: Kubernetes and CI/CD underneath, an LLM gateway and its guardrails on top.
Available from October 2026 · permanent or contract
Irish and British citizen — full right to work in Ireland, the UK and the EU, no sponsorship needed
devlinops
01 · case studies
Shipped and running
Six things I built and still look after. All but the last were at Loweconex, a UK IoT platform business in Northern Ireland.
Clarity
2025–26Natural-language database interface
Built the infrastructure and trust layer · Loweconex
Customers who wanted a number out of their estate used to raise a ticket and wait for an analyst. Now they ask directly — about twenty people a day, across roughly thirty tenants.
AI Gateway
2026One endpoint for every model
Built it and still own it · Loweconex
Adding an AI feature stopped meaning a new provider key and a billing conversation. It's a config change now, and spend traces to the tenant and feature that caused it.
Heimdall
2025Where every ticket and service actually is
Built and run it · Loweconex
The team stopped pasting kubectl output into Teams to ask whether a deploy had worked. Standup got shorter, and release management reads the same screen as the engineers.
Pipeline Platform
2024–26Shared CI/CD library
Built the shared pipeline library · Loweconex · still in use
A change to the build pattern used to mean a PR to twenty repos, so it didn't get made. It now ships once, and every service adopts it by bumping a tag.
Observability Stack
2024–25Self-hosted monitoring
Built and run the stack · Loweconex
Monitoring quoted near £100k a year runs in-house for about £5k, and two years on an incident starts with someone pasting a Grafana link.
Smart Home on K3s
2024–26Self-hosted home automation
Personal project · my flat
If the internet goes down, the lights still work — and nothing in the flat is reachable from outside the VPN.
02 · delivery
How my platforms ship
The pipeline takes it from here: build, test, scan, push the image. Nobody babysits a deploy.
The suite lives in its own repo and runs against a real environment. Tests are not pipeline steps.
Image Updater bumps the tag and ArgoCD reconciles the cluster to match.
Rollback is a revert, and the audit log is git log. I've rolled a release back at 2am with git revert and gone back to sleep.
03 · capability
What I build
Everything here is something I run in production or at home.
Kubernetes & EKS
EKS · K3s
Cluster operations end to end: node groups, zero-downtime upgrades, right-sizing for cost. The same discipline runs my single-node K3s homelab.
GitOps with ArgoCD
ArgoCD · Image Updater · git
ArgoCD and Image Updater reconcile the cluster to whatever git says, so nobody deploys by hand and the audit trail is the repository history. Everything downstream is derived state.
CI/CD pipelines
shared pipeline library
One shared pipeline library imported by every service: build, test, scan, push. ~400 deploys/month across 20 services on a single .ci/builds.yaml.
Observability
Prometheus · Grafana · Loki
Self-hosted Prometheus, Grafana and Loki. 22 dashboards managed as code, 50+ alerts with a runbook each, across four environments.
Security automation
SAST · SCA · policy
Veracode SAST and SourceClear SCA wired into CI so scanning isn't optional, with findings filed straight to Jira. Falco and Suricata I've used, but not carried a pager for.
Data platforms
Kafka · MSK
Kafka for service-to-service events, and the schema discipline that stops a producer change breaking every consumer downstream.
AWS
EKS · VPC · IAM · MSK · RDS
EKS, IAM, VPC, Route 53, S3, MSK, RDS. The boring fundamentals, done properly.
AI agents doing platform work
security triage · incident response
Agents pointed at the platform's own operational load rather than at a chat box: turning security findings into tickets, and taking the first pass on an incident against the runbooks that already exist.
Hiring for a platform team?
I'm most useful to teams putting AI workloads on Kubernetes, or to platforms that need CI/CD and observability sorted properly. I'm available from October 2026, permanent or contract, in Dublin, London, Amsterdam or remote-first EU. I usually reply within a day.