Case Study


Hypermode uses Axiom to win fans for their AI developer platform

Customer Case Study Screenshot
Ryan Fox-Tyler

SVP of Products and Engineering, Hypermode

Hypermode logo

Hypermode is a platform that makes it easy for developers to add AI to their apps.


  • Hypermode uses Axiom’s APIs and filters to let customers add messages and breakpoints from their own code. Hypermode then collates messages from separate executions that are exposed to customers.
  • As former users of large in-house tools, they knew building their own logging system would be far too challenging, and would not be a differentiator for customers. By adapting Axiom instead, Hypermode can focus their energies on rapid response and analysis of customer issues, which wins fans for their growing startup.
  • Unlike Splunk, Axiom allowed Hypermode to get started without an advance design and plan for indexes. They can just send everything to Axiom now and figure out how to structure it later.

About Hypermode

Hypermode, founded in 2023, is a platform that helps developers integrate AI features into their applications. Developers use it to create and handle AI models, functions, and data. It’s possible to instantiate a model with one line of code, then begin tailoring it with zero upfront data. They can focus on building their app without needing to become data science experts.

As an innovative new startup, Hypermode has 11 engineers who are encouraged to be full-stack. They recognized that customers would expect to produce logs from user-generated code and be able to monitor and analyze these events themselves.

With experience using both Splunk and custom-built solutions at previous companies, Hypermode engineers knew they wanted something less mammoth than Splunk, and hopefully with less upfront planning and work required to get the right data indexed usefully. They knew they wanted to focus on building their business, not building a logging platform.

“Axiom feels much more like a building block of your application rather than an internal tool.”

Ryan Fox-Tyler

SVP of Products and Engineering

The challenge: Let users generate and study their own logs

Hypermode at heart is a core runtime system where code interfaces with AI models and data sources, plus a model hosting service. The system produces logs from user-generated code, building up to a large, distributed system.

Engineers had two separate critical uses for a logging system: First, they needed to capture internal logs to spot and understand things that go wrong, and to support their customers.

Ryan Fox-Tyler, Hypermode’s SVP of Products and Engineering, had previously been vice president of the development platform at multinational insurance company Manulife, which holds over a trillion dollars in assets. He was also VP of products at Astronomer, makers of the rapidly growing data orchestration platform Astro. He had acquired plenty of experience with both Splunk and custom-built solutions.

What Ryan came to learn at those companies was that centralized logging was prohibitively expensive and hard to set up and maintain. The systems they used had come later in those companies’ lifecycles, when they could afford the cost and commitment. With Hypermode, he had joined a fast-moving startup that couldn’t yet afford the time or money Manulife and Astronomer had put into logging.

Second, Hypermode engineers wanted to expose relevant events to individual users, including the events produced by their user-generated code. That’s more complicated than it might sound. “The expectations of different users can vary wildly from one another, even for initial load time,” Ryan says. “I always felt like we were putting a square peg into a round hole with tools not really built for exposing to users in a way that feels native.”

The solution: Easy to set up, easy to expose to users

From the start, they found Axiom quick to set up — just following the docs for once worked, unlike a few others they had tried. Axiom’s omnivorous ingest and storage, plus the schema-on-read ability of its APL query interface, meant they could feed it everything now and then figure out as they go how they want to extract structured data from what they’ve collected. Axiom’s simple and affordable pricing also let them ingest everything without having to constantly monitor costs carefully.

To understand their customers’ use of Hypermode’s features, the team collect a high volume of Kubernetes logs. They also collect CloudWatch logs in Axiom to provide complete coverage over VPC and S3 activity, which provides a dependable audit trail.

Customers can add messages and breakpoints inside their own code. Hypermode not only collects them into Axiom, but collates messages from separate executions so customers needn’t pore through a long timestamped stream to find, for example, where the same button was clicked in dozens of separate runs. That makes analysis and debugging much, much easier.

Ryan says Axiom is the first logging product they tried that felt designed and built to capture and share customers’ logs individually. “You can expose them natively to your application,” he says. “Axiom feels much more like a building block of your application versus an internal tool.”

“Being able to actually go fix an issue much faster — ‘That fix is already in production’ — that’s how you win fans.”

Ryan Fox-Tyler

SVP of Products and Engineering

└ Hypermode’s UI displaying log events stored in Axiom.

Every engineer on the team uses it

Unlike Grafana, which Ryan found required a full-time engineer even in their cloud service to do proper curation, Axiom "lowers the barrier to creation and contribution by anyone who uses it.” Before giving up on Grafana, he didn’t see anyone other than the designated engineer contributing to it. By contrast, he says, “All of my team logs into Axiom.”

Hypermode engineers found it easy and quick to begin using Axiom for monitoring and analysis without facing a big learning curve. “I tend to steer away from these master-built dashboards that everyone thinks must be perfect, but they’re afraid to touch,” he says.

Instead, his engineers incrementally add to their Axiom dashboard as new issues show up. One function they’re working on is the ability to tie a trouble ticket in Plain to the logs in Axiom associated with its creation time, so that later investigators can immediately see what was happening in Hypermode’s infrastructure and the customer’s code at the time the problem was reported.

Logging won’t win customers, but better service will

Ryan’s executive view on logging is that for Hypermode customers, “it’s a high-value expected item. We’re not going to stand out from the market because we have logging. But being able to actually go fix an issue much faster — ‘That fix is already in production’ — that’s how you win fans.”

Interested to learn more about Axiom?

Sign up for free or contact us at to talk with one of the team about our enterprise plans.