Aidan Holsgrove Get in touch

Platform engineering Site reliability AI systems in production

Hi, I'm Aidan.

I build and run reliable platforms, and make it easier for engineers to ship software with confidence.

I've spent more than 16 years working across cloud infrastructure, production systems and developer tooling. Most recently, I led a distributed SRE team at Lightspeed, with engineers across Europe, North America and South America. I stayed hands-on with large production Kubernetes environments, data-intensive services, code and incident response.

I enjoy building practical AI tools that give engineers safe access to the context behind their code, helping them understand systems, investigate problems and improve what they build.

What I work on

Platforms & delivery

Cloud modernisation on AWS and GCP, particularly production Kubernetes clusters and services backed by large datasets. I've worked on database migrations and built safer release paths using GitOps, CI/CD, canary deployments and feature flags.

Reliability & operations

SLOs, observability, production debugging and incident response. I care about alerts that reflect real customer impact and tools that help an engineer understand what went wrong.

Practical AI for engineers

I built an AI incident investigator and a read-only MCP gateway that brought operational tools into engineers' existing AI workflows. Useful AI should make people more effective without giving up visibility or control.

A few examples

  • Safer releases. Built an isolated frontend canary across Go services, Helm and Gateway API, supporting cookie-based testing and percentage-based rollout while keeping the stable release separate. Used feature flags to control exposure and make changes easier to reverse.
  • Less on-call noise. Helped reduce monthly PagerDuty alerts from 199 to 45 by improving alert quality, ownership and runbooks.
  • Faster delivery. Helped cut CI pipeline time from 18 to 5 minutes through CI/CD and GitOps improvements.
  • Infrastructure modernisation. Migrations I've led or delivered with my team:
    • Web platform from on-premises hosting to AWS
    • Applications from Mesos to Amazon EKS
    • Kafka to Amazon MSK
    • Elasticsearch on EC2 to Amazon OpenSearch
    • Self-hosted MongoDB to MongoDB Atlas
    • Source control and CI from GitLab to GitHub

AI Agent 007

Agent 007 is an AI incident investigator I designed and built at Lightspeed. It picks up alerts from PagerDuty, incident.io or Datadog, gathers read-only evidence from Kubernetes, metrics, logs and recent deployments, then posts what happened, the likely cause, customer impact and a recommended next step in Slack.

Engineers could also talk to it directly. They could mention it in Slack to ask a question about the platform or investigate an incident, then reply in the thread to follow up.

It was adopted across six Lightspeed product stacks on GCP and AWS and investigated more than 300 incidents, usually reporting back within about three minutes. I also built a read-only MCP gateway so engineers could use the same tools in their own AI workflows, and my team built First Pass, which turns confirmed findings into proposed code fixes for human review and has completed more than 1,000 production pull-request review runs.

Illustrative incident summary with what happened, likely explanation, customer impact, next step and confidence level.
On-call investigation. An example of the evidence, impact and recommended next steps 007 brought into the incident thread.
Illustrative operations dashboard with cross-team investigation search, environment health and recent investigations.
007 HQ. A cross-team view for finding previous investigations, checking capabilities and asking questions about the platform.

Both illustrations are anonymised recreations inspired by internal tools, not screenshots of live systems. Incident details are simplified or fictional.

Background

At Lightspeed and ShopKeep, I worked across SRE, platform engineering and engineering leadership. I built and coached an SRE team, worked on Kubernetes and database modernisation, and helped teams improve how they deploy and operate software.

Before that, I worked at Corus Entertainment in Toronto, Canada, supporting the websites of TV and radio channels including HBO Canada, Nickelodeon and Teletoon. I built Corus's first autoscaling web platform on AWS, moving it off on-premises hosting.

Technologies I work with: Go, Python, JavaScript, Kubernetes, Terraform, ArgoCD, PostgreSQL, MongoDB, Prometheus, Grafana and OpenTelemetry.

Get in touch

I'd be glad to talk about platform engineering, reliability or developer tooling.

Connect on LinkedIn