Customer: WHOOP
Industry: Healthtech
Use Case: Data pipeline and machine learning orchestration
Key Outcomes:
WHOOP is a wearable that helps people listen to their bodies. It tracks metrics like sleep, strain, recovery, and stress, and delivers science-backed feedback through a personalized analytics platform. The device is sleek and screenless, designed to be worn around the clock and fade into daily life.
Carlos Peralta, WHOOP's Director of Data Platforms and MLOps, explained the company's mission this way: "We want to help people become their best selves. That might mean optimizing performance for a triathlon, or just getting better sleep after a tough week. WHOOP fills the gap between how you feel and what your body's actually telling you."
While the experience feels simple to the user, the technical systems behind it are anything but. WHOOP builds its own hardware, develops its own software, and processes a staggering volume of data to generate insights in near real time. Getting the data right is essential. "We're a data-driven company," Carlos said. "If we deliver the wrong insights, we lose trust. And when you lose that, it's hard to earn back."
When Carlos joined WHOOP, data workflows were being orchestrated through a collection of internal tools. They worked, but not well enough. As use cases grew, so did the pain. "We were relying on homegrown systems, but everything felt duct-taped together. We didn't have the observability we needed. When a job failed, we had to dig through logs or ask around to figure out what broke."
The tooling was hard to scale and hard to debug. "Scaling across use cases was very challenging," Carlos said. "And when we had to scale, it usually translated into a lot of overhead."
WHOOP's workflows support internal analytics, machine learning, and member-facing features. The risk wasn't just slow pipelines, it was broken experiences for users expecting reliable health insights.
Carlos's team started with the free tier of Prefect Cloud. "The free tier gave us a hosted control plane and observability out of the box. That was enough to move quickly," he said. They began by orchestrating lightweight Python ETLs, moving data from internal sources into Snowflake. From there, adoption spread quickly.
Over time, Prefect became the foundation for WHOOP's orchestration layer. The team now uses it to coordinate data and ML workflows, including:
The real turning point wasn’t just adding more flows, it was seeing how Prefect handled growing complexity through dynamic orchestration without getting in the way. “We found that after a very light POC, it fit very well,” Carlos said. “The scalability was easy to manage, and the overhead to add new workloads was very low.”
As their workflows scaled, so did their requirements. "We needed SLAs, audit logs, more concurrency, and real support," Carlos said. As orchestration became more central to WHOOP's production systems, the team needed stronger guarantees and governance. "We were orchestrating jobs that touched product experiences and machine learning workflows. It wasn't just about running flows anymore, it was about control, observability, and uptime."
The shift from homegrown tools to Prefect brought measurable improvements. "After switching to Prefect, the incident count was cut by 75 percent," Carlos said. "Our mean time to recovery improved by over 40 percent thanks to the observability we get with Prefect."
More importantly, engineers got time back. "We've been able to cut orchestration overhead by days per sprint with Prefect," Carlos said. "If you think about a sprint of 10 days, and you can cut 2 to 3 days of that workload per engineer, you're cutting 30 percent of their time."
This shift unlocked broader benefits:
WHOOP's infrastructure runs in AWS, with workflows deployed in Kubernetes. "Prefect flows can access our AWS resources through IAM and Kubernetes pod-level identities," Carlos explained. "We control everything at the Prefect Work Pool level."
One early challenge was managing secure access across services while scaling orchestration. Prefect's support team helped them implement a robust model that worked with WHOOP's existing controls without requiring architectural changes. "We didn't have to rewire anything. Prefect fit right into the way we build."
Carlos credits the Prefect team for making the rollout a success. "From day one, the team gave us clear information. They told us, 'This is how it works. This is what it costs. This is how we'll support you.' And they've followed through."
He described the relationship as "a bear hug. We're going to be very close. Buying tools is easy, but what makes a partnership successful is what happens afterwards. And there's been no disappointment."
For teams evaluating orchestration tools, Carlos advises: "Don't overthink it. As engineers, we always want to overengineer everything. But sometimes, the best tool is the one that gets out of your way."
He suggests starting with a small POC. "Just try Prefect. Let a couple of people run their workflows with it. You'll see the difference right away."
WHOOP is continuing to scale its data platform with Prefect. The team is migrating configurations to Terraform, setting up separate environments, and standardizing observability across workflows.
"Prefect gives us the reliability we need and the speed we want," Carlos said. "It's part of how we scale, and it's part of how we stay fast."
Curious how companies like WHOOP, Cash App, and others build resilient data platforms with Prefect? Get started for free and experience it firsthand.
