Spacelift’s Flexibility Unlocks Faster Scaling for Duolingo Within Months

Duolingo is the most popular mobile learning platform and the most downloaded education app worldwide. The app makes learning new languages, math, music, and chess fun with bite-sized lessons that feel like playing a game. The company’s mission is to develop the best education in the world and make it universally available. Duolingo offers lessons in music, math, chess, and language with over 250 total language courses from Spanish, French, German, and Japanese to Navajo and Yiddish. 

Evan Strat, Senior Platform Engineer at Duolingo, spoke to us about the ways Spacelift’s flexibility has allowed the company to scale its infrastructure safely — and fast. 

Before they adopted the Spacelift platform in October 2024, the Duolingo team managed Terraform with Atlantis, but it was difficult to keep Atlantis stable at scale and implement new features. 

Some specific challenges included:

• PR lock workflow slowed things down painfully.
• Limited support for merge-before-apply workflows 
• Inability to scale past four concurrent plans and one concurrent apply
• Issues with storage when running Atlantis on ECS in a standard microservice format
• Difficulty writing Terraform policy
• Problems using Atlantis with multiple AWS accounts

“We chose Spacelift because we felt it had the most robust platform and would offer the best experience to both our platform team doing most of the configuration in Spacelift, and our engineers who only lightly touch it,” explains Evan. As Duolingo scales its IaC, it needs a platform it can rely on. Less than six months after its adoption, Spacelift had already proved its worth. “Spacelift is robust, reliable, and customizable. There’s a lot less babysitting and worrying; it’s a lot smoother.”

Duolingo had been finding workarounds for some of its Atlantis issues — for example, adapting it to use a persistent file system to counter storage issues — but trying to scale with Atlantis was hugely labor-intensive. As Evan explains, “We had to update Atlantis versions manually, but with Spacelift I have not thought about how it actually runs Terraform in several months because we tuned our worker instance size and it just hasn’t been a problem since — and that frees us up to solve actual interesting problems!”

“Spacelift provides the right kind of hooks for us to add our customizations without having to worry about the core Terraform.”

Evan Strat, Senior Platform Engineer

As the Duolingo team expands its Spacelift experience, they are discovering the ease with which they can adapt the platform to the way they do Terraform. As Evan Strat concludes:

“If you need to deeply customize your Terraform workflow, Spacelift is a great option. Duolingo has some very particular Terraform workflows, and we’ve been able to perfectly customize Spacelift to fit our needs.”