Lead Platform Reliability Engineer,
Nov, 2020 - Present
My role as Lead Platform Reliability Engineer is primarily a technical leadership and individual contributor role, with no line management. I work across engineering rather than being restricted to the platform team, usually splitting my time between hands-on engineering and architecture, technical direction, mentoring and helping other teams work through difficult problems.
A large part of the role has become being someone teams can bring an ambiguous or particularly difficult problem to. That can involve infrastructure, application architecture, performance, event-driven systems, CI/CD, security or simply another perspective on a decision. I enjoy getting involved where the boundaries between teams and technologies become less clear, particularly when the solution requires understanding the problem rather than starting with a particular technology.
I've repeatedly taken ownership of technical direction where there wasn't already a clear roadmap, including our multi-account AWS migration, ephemeral account platform, CI/CD evolution and internal AI platform. I tend to build these as self-service patterns rather than centralised gates, so the safe and supported route becomes the easiest route and teams can keep moving without needing the platform team involved in every change.
More recently I've been leading much of our exploration around practical AI usage in engineering. This has included evaluating models, coding agents, harnesses and context strategies, building tooling to understand how engineers actually use them, and working out where agents add real value without weakening engineering standards.
Independently conceived, built, released and continue to maintain an internal AI application platform used across the organisation. It allows non-technical users to create persistent applications through a conversational interface or MCP while automatically handling Okta authentication, permissions, data storage and security controls, including user-aware data isolation backed by PostgreSQL row-level security.
The internal AI platform now hosts more than 50 applications, ranging from security training and engineering visibility tools through to product launch and inventory readiness systems. It has also recreated and automatically populated an internal event catalogue by indexing more than 500 Git repositories.
Own the roadmap and scope for the internal AI platform, deliberately keeping it focused on current business value while preserving future ideas rather than allowing the project to grow into everything it could theoretically do.
Built deep analytics and evaluation tooling around agentic engineering, including harness and model comparison, context usage analysis and benchmarking, to understand what LLMs are actually doing and help engineers use them more effectively.
Help shape our wider AI-in-engineering roadmap and lead practical adoption of AI tooling, including AWS Bedrock guardrails and secure access patterns, while keeping engineering judgement, data protection and organisational controls in the loop.
Designed and led our engineering migration from Codefresh to GitHub Actions across roughly 300 repositories. I built more than 50 reusable actions and around six composable workflows that teams can adopt directly or extend themselves.
Redesigned the CI approach around build-once-and-promote principles, merge queues and repository-managed AWS OIDC access. Some of our largest pipelines reduced from more than an hour to under 15 minutes as part of the migration.
Turned platform decisions into self-service organisational defaults. Patterns such as repository-managed AWS OIDC, shared CI actions and workflows are used directly by teams with minimal platform involvement, while required controls can be enforced without removing team ownership.
Worked closely with our web engineering teams across several generations of the Gymshark ecommerce platform, including migrations from Shopify to Kubernetes, Kubernetes to AWS Lambda, a later multi-region Lambda architecture, and the current work to split dependencies and functions into a more modular serverless design.
Led and supported performance testing and architecture work for major launches including Black Friday and Onyx, where the primary website needs to move almost instantly from roughly 400 requests per second to more than 100,000 requests per second from a cold start, without prewarming or application-side degradation.
Before one Black Friday event I identified an uncached and unusually expensive 404 path that could cause the entire website to fail during a burst of missing content. I reproduced the failure mode, linked it to previous publishing incidents and worked with the team to address it before launch.
Introduced Datadog at Gymshark and remain one of the primary technical contacts for its architecture and usage, including logs, metrics, distributed tracing, OpenTelemetry conventions, data protection and cost control.
Identified and implemented infrastructure and observability changes representing more than £1m per year in avoided expenditure, including database rightsizing, Kubernetes compute optimisation and reducing unnecessary metrics and logging spend.
Defined and drove the roadmap for Gymshark's move to a multi-account AWS organisation as it grew to more than 60 AWS accounts. This started with fixing identity and access, then identifying the target team/environment boundaries, sequencing migrations and ultimately decommissioning the legacy account model.
Designed much of Gymshark's AWS identity model, replacing per-account usernames and passwords with central Okta SSO, organisation-wide auditing and logging, and automated team and environment provisioning.
Built the original ephemeral AWS account workflow that allowed pull requests to provision isolated AWS accounts, deploy infrastructure, grant the required access and clean everything up when the work was complete. When AWS later released Innovation Sandbox I worked with AWS on its adoption and replaced our bespoke implementation rather than continuing to maintain something the platform could now provide directly.
Built and maintain a Go client for AWS Innovation Sandbox and a GitHub integration that provisions sandbox accounts and deploys Terraform from pull requests.
Introduced feature toggling across engineering using a modified Unleash setup, giving teams a shared way to separate deployment from release without introducing the cost and complexity of a larger commercial platform we did not need.
Helped shape Gymshark's event-driven architecture over several generations, including the original event-sourcing approach and work involving RabbitMQ, EventBridge and Kinesis, with practical experience around queue failure handling, retries and backoff. I am regularly asked to review or sign off event flows and AWS architecture outside my own team.
Remain deeply involved in major incidents and difficult production investigations, including application outages, infrastructure failures and sensitive data incidents where I've worked across engineering, security and legal teams to establish technical cause, impact and appropriate remediation.
Write and maintain technical and user-facing documentation for platforms I own, including Innovation Sandbox, account factory automation and the internal AI application platform.
Mentor engineers individually and through technical sessions, regularly advise five core engineering teams as well as application and platform teams, and have participated in hiring engineers, managers and directors. People across engineering regularly bring me complex or ambiguous work for challenge, adversarial review or an outside technical perspective.