How Primer uses ML to build safe online communities for kids

“We would be behind if it weren’t for Baseten. The biggest reason we chose Baseten is its ability to scale. I know Baseten will be able to keep up with our growth, and we can augment what we’re doing on the platform whenever we need to.”

Ryan Delk, CEO & Co-Founder, Primer


Growing up homeschooled, Ryan Delk, CEO and co-founder of Primer, experienced first-hand the benefits a non-traditional education could provide kids. Believing that every kid should have access to the experiences he had, where learning environments are tailored to their unique gifts and interests, he co-founded Primer.

Primer’s first product, called Clubs, is a series of interest-based communities designed for students. Kids can join communities that align with their interests, where they can collaborate on projects and learn alongside like-minded peers.

Having raised $3.7M in seed funding in 2020, Delk says Rooms is just the start for Primer’s rapidly growing user base.

The Problem

As a part of building online experiences for kids, ensuring their members’ privacy and safety is always top of mind. “Anytime you’re building a product—but especially when you’re building one for kids—keeping them safe is the highest priority,” said Delk.

One of the key challenges the team needed to solve was verifying that the users who signed up for Primer were actually kids. “We needed something that was both seamless for the user experience, but also highly robust and would give us data that we’re really confident in,” said Delk.

Searching for a solution, Primer’s team were presented with the classic dilemma: build in-house or buy a third-party solution. For Delk, neither option was optimal.

The time required to build something in-house meant the already lean team would need to make sacrifices on new feature work that would directly benefit users. But purchasing a productized solution would not only be expensive, it also would mean they’d be constrained by the limitations and product decisions of that provider. 

The Solution

Hoping to find another way, Delk was introduced to Baseten. Using Baseten, the team could quickly deploy their user verification ML models and integrate them into Primer’s user-facing application—all with just a few lines of Python. When a user required additional verification, the team used the drag-and-drop Views to build an internal app for Primer’s team to review and approve one-off cases.

All of this was accomplished without worrying about any of the backend, frontend, or MLOps engineering that would typically be required, enabling the team to move much faster than they had initially thought possible. “Baseten allows us to take any need that we have and quickly build a custom SaaS solution for that very specific use case. It’s amazing,” said Delk.

In addition to moving faster, building with Baseten meant the team didn’t need to sacrifice any of the fine-grained control they wanted in order to provide a seamless, safe experience for their users. “Baseten is modular by nature, which allows us to build on top of it and make it feel like a native part of our system,” said Delk.

The Result

With Baseten, Primer’s team had a working user verification solution in less than three weeks. “It was the fastest deployment we’ve ever had,” said Delk. 

Had they decided to build everything from scratch, Delk estimates they would’ve needed to staff two engineers full-time for 3-6 months, not taking into account all of the ongoing maintenance and upkeep. Instead, “our team is working on new features to prepare for a product launch in a few weeks,” said Delk. “We would be behind if it weren’t for Baseten.”

Looking ahead, Delk sees Baseten as a long-term partner for Primer. “The biggest reason we chose Baseten is its ability to scale. I know Baseten will be able to keep up with our growth, and we can augment what we’re doing on the platform whenever we need to.”