TGL Golf — Game Operations Platform

01 The Challenge

TGL — the indoor golf league co-founded by Tiger Woods and Rory McIlroy, played in a purpose-built simulator-driven arena — needed a bonafide Kubernetes architect to stand up the platform that runs live matches. Each match relies on 16 distinct game components, each responsible for a slice of the simulator gameplay, broadcast lighting, scoring, and operational telemetry that powers a live televised event.

The 16 components had to deploy reliably onto Red Hat OpenShift in two very different environments: a development cluster used for iteration and rehearsal, and the on-prem Kubernetes cluster physically running inside the Florida arena on game day. With a live broadcast on the line, “works in dev” was nowhere near enough — the platform had to behave identically the moment the cameras went live.

02 The Approach

Leopard Data led the project as the systems group — a three-person team responsible for architecture and delivery of the full CI/CD platform. The pipeline pulled source from each component’s repository, ran builds, packaged everything as OCI containers, and orchestrated rollouts of all 16 components into OpenShift. Each stage of the pipeline was instrumented for visibility so the operating team could trace any deploy from commit to running pod.

Infrastructure provisioning was modeled in Red Hat Ansible Automation Platform, giving us a single declarative source of truth for both the dev cluster and the on-prem game-day cluster. Environment differences — networking, storage classes, secrets, hardware profiles — were handled through parameterized Ansible playbooks rather than divergent ad-hoc scripts. That dev/prod parity discipline is what kept the game-day deploy boring instead of terrifying.

The CI/CD platform itself was delivered to Full Swing Golf — TGL’s simulator partner — as a working artifact, so their engineering team could continue iterating on the game stack with the same workflow Leopard Data established.

03 The Delivery

Pre-game-day verification happened on a sound stage at Universal Studios in Orlando, wired up to mimic the production environment as closely as possible. The full pipeline was exercised end-to-end against a representative cluster, and operational runbooks were dry-run with the team that would actually be running the system on game night.

Once the dev pipeline was rock-solid, Leopard Data traveled on-site to the TGL arena to re-adapt the Ansible automation for the actual on-prem Kubernetes cluster physically running the game. That meant accounting for the constrained network topology of a stadium, the specific on-site hardware profiles, and the broadcast-grade reliability bar that a one-shot live event demands.

05 Tech Stack