Migrating production infrastructure? Get up to $10K in migration credits.
Apply nowHow Mixtiles decorates homes, not infrastructure
"We want to spend less time on DevOps-level infrastructure and more time on how to make your home experience the best in the world," says Chen Cozocaru, Mixtiles' Infrastructure Engineering Manager.
Mixtiles builds photo tiles, photo books, AI-generated templates, self-watering plants under the EasyPlant brand, and smart lighting, products meant to make a house feel more like home. It is backed by an algorithms team that generates photo books automatically from a customer's photo feed. About 100 people run the company, with no in-house DevOps function, and seasonal demand that can run 3x, 5x, and sometimes 10x normal volume around Black Friday and Cyber Monday, when a few hours of downtime can cost the company millions.
A product team that size, with no DevOps function of its own, needs its infrastructure to just work.
The dashboard went down too
Mixtiles had run on their previous hosting platform for more than a decade, largely because it let their small, product-focused team stay out of the infrastructure business. Then an outage took the application down, and took the dashboard down with it, so the team couldn't see or control what was happening in real time.
With peak sales hours rapidly approaching, Mixtiles’ team split up and started testing paths in parallel to get the online order flow back up. "We're a very successful B2C company, so every day we're down is a lot of revenue lost," says Cozocaru. They tried Vercel first, which had a technical gap with their SSR rendering setup, a variation of Remix, and AWS Elastic Beanstalk, which was slower to get running.
Render was the first option that got the stack live immediately. Within about five hours, Mixtiles had the online ordering path running on Render, enough to keep taking customer orders before the backend and fulfillment systems had moved over.
In Cozocaru’s experience, Render’s support was equally quick and reliable: "It’s the highest I've ever seen. The tech experts were really in the details, and they had really good advice."
That support continued well past the initial outage. Cozocaru says the team could raise a question with Render’s account team in Slack, even a non-urgent one, and count on getting a real answer back quickly. "It's not a shallow answer. It's an incisive answer with content," he says.
With the emergency behind them, Mixtiles turned to a harder question: whether Render was the right long-term fit, not just the fastest one.
From a short-term fix to a long-term fit
The Render setup that got them through the outage had been built fast, under pressure. It was effective as a quick fix, but not yet the kind of structured production setup they wanted to run long term. So instead of forcing a rushed decision, they took the time to make the call.
The decision hinged less on the outage itself than on whether Render fit how Mixtiles wanted to operate: a small, deeply product-focused engineering team that wanted to confidently handle sharp seasonal swings without spending time on infrastructure.
A second incident put Render's fit to the test. An AWS disruption affected a critical piece of Mixtiles' production-order system that hadn't moved off their previous setup yet, and it took about 12 hours to come back online. For a company running through seasonal spikes, that kind of downtime doesn't stay contained to a dashboard, it delays real orders. Rather than wait, the team ran the same workload as a one-off job on Render, where it was running again in roughly 30 minutes. "It was proof that Render was the right direction," says Cozocaru.
Six months, then one week
Mixtiles moved to Render one service at a time, keeping production running throughout. The work included adapting to differences in worker configuration, figuring out how to handle many cron jobs that shared a single repository, and reworking parts of the deployment pipeline around Render's model for previews and services. Over roughly six months, Cozocaru says he "didn't encounter any showstoppers," and his original estimates for the timeline held up.
Later migrations went faster. "We thought it would take like two weeks to migrate to Render but it only took one week," Cozocaru says of Easyplant, another product they operate.
What staying lean gets them
Mixtiles now runs its online ordering flow, cron jobs, and infrastructure for a growing list of additional services on Render. Preview environments matter most for the online side of the business, letting the team test changes against something close to production before they ship. Easyplant alone has scaled to 40 to 60 servers at times, according to Cozocaru, still far from what he thinks is its potential peak.
Cozocaru estimates Render runs about 40% cheaper than what Mixtiles paid before. And as Mixtiles has grown, the team has brought new services onto the platform. Setting up a new service got faster too. "Starting the service in Render is really fast, so it's probably like 80% faster than before," Cozocaru says.
That mix of savings and speed is why Cozocaru is confident Mixtiles can keep scaling on Render. "Render is totally production grade," he says.
AI-first, but never on autopilot
Mixtiles' product-first mindset shows up in how the team writes software, too. Cozocaru says about 99% of the company's coding is AI-assisted, primarily with tools like Claude and Codex, and that the pattern has spread beyond engineering into other teams as well. "The number one challenge for me now is to use it and not treat it as a black box," he says.
Scaling a product-first company
That same instinct runs through the algorithms team, which builds features like automatically generated photo books from a customer's photo feed, the kind of product work that depends on engineers spending time on the product and not on keeping servers alive.
As Mixtiles adds brands like Easyplant and leans further into AI-assisted development, the goal is still to spend less time on infrastructure and more time making their customers’ home experience the best in the world.
