Platform engineering
Also known as: internal developer platform
Platform engineering is the practice of building an internal platform, a curated set of self-service tools, templates, and infrastructure, so product developers can build, deploy, and run software without becoming infrastructure experts.
While it's often pitched as the successor to DevOps, it's really a correction to it: 'you build it, you run it' turned out to overload developers with cloud complexity. A good platform team treats other developers as customers.
Platform engineering, explained at three levels
At home
Imagine a shared kitchen where someone has already stocked the pantry, sharpened the knives, and written simple recipes. You just cook. Platform engineering does that for software developers: it sets up the tools so they can focus on building apps.
At work
In a company, a platform team builds the internal tools developers use every day: templates for new services, automatic deployment, logging, and security checks. Developers click a button or run a command instead of filing tickets and waiting.
At home
The key idea is a 'golden path' or 'paved road': a recommended, well-supported way to do common tasks, like creating a new web service. You can go off the path if you really need to, but the path is so easy that most people don't.
At work
A typical internal developer platform combines infrastructure as code, CI/CD pipelines, a container platform such as Kubernetes, observability, secrets management, and a developer portal. Success is measured by developer adoption, lead time to production, and how much toil it removes, not by how many tools it includes.
At home
The core trade-off is abstraction versus flexibility. Hide too much and developers can't debug problems; expose too much and you've rebuilt the cognitive load you were trying to remove. In my opinion, the best platforms are 'thin': opinionated defaults with clear escape hatches.
At work
Treat the platform as a product: interview internal users, publish a roadmap, version APIs, and measure satisfaction alongside DORA metrics. Common failure modes include mandating the platform before it's good, building for hypothetical scale, and staffing it purely with infrastructure engineers who've never shipped a product feature. Funding should be tied to measurable reductions in developer toil.
Real-world examples
At home
Sophie joined a startup as a junior developer. On her first day, she created a new service from a template in the company's developer portal, and it came with logging, a deployment pipeline, and a test environment already set up. She shipped her first change before lunch.
At work
Thomas leads engineering at a Calgary insurance company where every team had built its own deployment scripts. He formed a small platform team that created one supported path for new services. Within a year, setting up a service went from two weeks of tickets to under an hour.
How it compares
| feature | Platform engineering | DevOps |
|---|---|---|
| Main idea | Build self-service tools for developers | Break down walls between development and operations |
| Who runs infrastructure | A platform team provides it as a product | Often each product team |
| Developer experience | Golden paths and a portal | Varies by team |
| Risk | Becoming a new gatekeeper | Overloading developers with operations work |
Platform engineering keeps DevOps goals but centralizes the tooling so product teams don't each rebuild it.
For analysts
Before funding a platform team, measure where developers actually lose time. I'd start with the two or three most painful workflows, usually new services and deployments, and prove value there before building a grand platform nobody asked for.
History & origin
DevOps spread through the 2010s with the idea that developers should own operations. As cloud and container tooling exploded, many teams found themselves maintaining Kubernetes manifests, pipelines, and security policies instead of writing features. The 2019 book Team Topologies, by Matthew Skelton and Manuel Pais, popularized the 'platform team' that reduces cognitive load for 'stream-aligned' product teams. Spotify open-sourced its Backstage developer portal in 2020 and donated it to the Cloud Native Computing Foundation that same year, giving the movement a flagship tool. By the mid-2020s, analysts were predicting most large engineering organizations would have platform teams. In my view, platform engineering succeeds when it's run like a product and fails when it's run like a gatekeeping ops department with a new name.