How OCMI moved beyond AWS to focus on building product
Founded
2012
Industry
Insurance
Migrated from
AWS
OCMI Workers Comp is a recognized leader in specialized insurance and earned Inc. 5000 honors in 2023. The company operates as part of a larger conglomerate that includes PEOPayGo and COMPEO. Multiple businesses within the group have earned industry recognition. PEOPayGo joined OCMI Workers Comp on the Inc. 5000 list in 2023. COMPEO was also named an Inc. Best in Business honoree in 2024. The proprietary Janus Platform powers the CRM, payroll, and onboarding for this entire multi-company ecosystem.
A lean engineering team manages this complex platform at the intersection of workers’ compensation, payroll, and HR services. At their scale, infrastructure failures are not just logs in a dashboard. These failures have real human impact. A silent webhook failure stalls a worker's payroll processing.
As Cristian Herrera, CTO at OCMI, puts it, “No one likes to not get paid.”
Accordingly, OCMI's engineering team had three non-negotiable infrastructure requirements: deploy and manage applications without DevOps overhead, scale intuitively and recover fast when something breaks, and work with a platform that offered architectural support and could grow alongside them.
AWS gave them control. What it didn't give them was simplicity, reliable scaling, or a real support partner. For a team running workers’ comp and employee payroll, that tradeoff was not working. The infrastructure had become the job. Finding a platform where that was not the case took a disciplined search. It ended with Render.
When infrastructure management replaces product building
AWS wasn't failing dramatically. It was failing quietly and constantly.
Broken builds took five minutes to recover with no easy rollback path, forcing the team to babysit CI pipelines. Autoscaling meant standing up load balancers and manual tuning; overprovisioning was the only way to ensure reliability without constant maintenance. And when they needed architectural input, AWS support responses arrived too late, rarely offering guidance on the specific decision at hand.
The code side of their stack was modern and getting faster. The platform they deployed it on was pulling in the opposite direction. As Herrera recalls, "it really came to a point where we thought, okay, we're going to have to hire someone for this. This is a lot."
For a lean engineering team whose job was building products, creating a function solely to maintain infrastructure wasn't a solution. It was a sign that they needed a more modern cloud.
OCMI offers Pay-As-You-Go PEO Programs as an alternative to traditional Workers Comp plans.
Evaluating the best architectural fit
OCMI didn't abandon AWS impulsively. They ran a disciplined evaluation across Render, Vercel, Railway, Cloudflare, and others to understand which provider could give them a fundamentally different relationship with their infrastructure; one where deployment, scaling, and recovery worked without specialist knowledge, where support was responsive and provided genuine architectural guidance, and where the platform could adapt as their business grew.
"Railway looks simple on the surface, but once you try to do anything real, the complexity is still there, especially around scaling. We'd still need to set up our own load balancer, specify how requests are routed to it, and work around runtime restrictions," explains Herrera.
"Vercel is a different story. If you need to do anything outside of Next.js, you're fighting the platform. Serverless was too narrow for our architecture. None of them found the right middle ground between customization and developer experience."
Render didn’t require these tradeoffs. Deployment was straightforward without runtime restrictions or manual load balancer setup. Autoscaling was intuitive and reliable. And for the first time in their search, they felt like they had a real partner.
Migrating in a week
As OCMI modernized its application stack, the underlying architecture accumulated the kind of sprawl AWS tends to produce over time: EC2 instances managed by Laravel Forge, RDS databases, ElastiCache for Redis, Terraform and AWS CDK for infrastructure-as-code, plus CloudWatch, SES, SNS, and Route53. The application code lived in a poly-repo structure spread across all of it.
Post-migration, Render handles most of OCMI's previous AWS services natively. Services are grouped by project in a monorepo and deployed via Render Blueprints, each containerized and independently autoscalable with a single toggle. Redis runs as a single right-sized instance rather than a cluster, PostgreSQL serves as the primary database for migrated services, and applications were refactored from PHP to Node/TypeScript in parallel with the infrastructure move.
“The migration itself was unremarkable, in the best sense,” according to Herrera. “We could migrate most applications within a week.” Even though OCMI was simultaneously refactoring into a monorepo, a task that usually invites chaos, Blueprints and Render’s API made it a quick move with nearly zero downtime.
Throughout, Render wasn't just a vendor, it was a working partner.
"We really saw Render put in the effort to understand our applications, to understand even things more abstract like our vision. They were actually trying to help us solve the problem. They were on the front lines with us."
“We were using agents to make writing code 80% faster, but the moment we went to deploy something new, we were back to spending hours on AWS configuration. The infrastructure just wasn't keeping up with how we were building.
”
Hiring a dedicated DevOps engineer to manage infrastructure had felt inevitable on AWS. On Render, it has never come up again. OCMI never made that hire and that absence became one of the clearest signals the migration had worked.
Following the migration, recovery time is down from five minutes to five to ten seconds. “We’re also on track to save 35 to 40% on compute vs AWS because autoscaling is so straightforward,” notes Herrera. These aren't enterprise discounts. They're raw compute savings from finally only paying for what they actually use.
The operational shift has shown up in quieter ways, too. Any engineer on the team can now deploy, set environment variables, and understand what is running, all without deep infrastructure expertise. As Herrera puts it: “The platform is intuitive enough that the team rarely needs to ask for help at all.”
Downstream, the outcomes OCMI actually cares about are moving in the right direction. Employees are onboarded without disruption, payroll events are processed reliably, and workers' comp claims are handled on time.
Spending engineering time on what matters
The deeper bet OCMI has made isn't just about tooling. It’s the conviction that the best engineering organizations of the next decade won't be the ones with the most bespoke infrastructure setup, but rather those who've cleared that path entirely, so engineers can focus on what actually matters: building product.
