Monolith to Micro-Services
Replaced a large server with frequent breakages, and built a subscription engine with zero breakage that handles massive traffic.
PetLab Co. was already the world's fastest-growing pet supplement brand: bootstrapped, profitable, 9-figure. What got them there wouldn't get them where they wanted to go. We rebuilt the platform piece by piece, then re-architected the marketing engine around generative AI.
Bootstrapped to nine figures by 2022, but the platform and marketing engine that built the company
weren't the ones that would scale it.
So we rebuilt both.
PetLab Co. was established in 2018 and became the world's fastest-growing pet supplement company: nine-figure revenue by 2022, bootstrapped, profitable, and still moving fast.
The growth created two problems: a platform that kept breaking and a marketing engine designed for a smaller scale. Throughout 2023, AI Performance Lab (led by Dr. Michael Housman) led a comprehensive re-architecture of PetLab Co.'s platform, then turned the marketing engine into something that compounds, not something that scales linearly with hires.
By late 2022, the platform was straining under more than 1.5 million dogs and their owners. It had been designed for a fledgling startup: a monolith on oversized servers, a patchwork of languages and frameworks, and serious security gaps. It broke often, and every outage cost revenue.
An online store can't close for six months, so we rebuilt it one piece at a time while the business kept running. Dr. Michael Housman served as interim CTO, with a weekly tiger team across frontend, backend, QA and DevOps, developers from AI Performance Lab, and PetLab's AWS account team.
Replaced a large server with frequent breakages, and built a subscription engine with zero breakage that handles massive traffic.
Moved to React and Next.js with shared components. Google page-speed scores rose from 13 to 50-60, which lifted conversion.
Replaced manual configuration with infrastructure as code, canary deployments and 500+ automated tests.
Added SSO, logging and full web application firewall coverage. The AWS security score doubled.
A rebuild pays off late, so we looked for wins along the way. Subscription retry was one of the best. When a subscriber's payment fails, the system tries again. Monthly revenue from those retries sat around $55K through 2022. It nearly doubled to $105K by June 2023 and reached $230K by January 2024, for $2.0M recovered in total.
"During this period of re-architecture, not only did we continue to grow significantly, but we also successfully rebuilt our entire platform. Now it stands more scalable, stable, and secure than ever before."
Damian GrabarczykBusiness Manager, PetLabCo
Four constraints stacked on top of each other: each one tolerable on its own, all four together a hard cap on growth.
Slow generation of diverse and engaging content: couldn't respond to real-time marketing dynamics.
Traditional content workflows limited the team's ability to produce innovative ads that resonated across diverse audiences.
Excessive time and resources spent on production, pulling marketing away from strategy and analysis.
Couldn't scale campaign volume without proportional increases in headcount and overhead.
Marketing becomes a closed loop: create → launch → gather → create. AI-driven creative generation produces on-brand variants at production scale; campaigns launch; performance data and audience feedback come back in; the next round of creative is informed by what just worked. The loop runs continuously.
AI-driven creative generation produces on-brand ad variants (product imagery, lifestyle scenes, copy) across every format the team ships.
Campaigns ship faster, in more variants, across more audiences. Production stops being the bottleneck.
Performance data and audience feedback flow back in: what landed, what didn't, on which segment.
Next round of creative is generated against what the data just said. The loop closes. The team's job becomes the loop, not the artifacts.
AI doesn't replace the marketer: it removes the part of the job that scales worst.
Two charts tell most of the story: campaigns got dramatically cheaper to produce, and the team shipped dramatically more of them. Revenue followed, not in lockstep, but with the lag you'd expect from a media engine warming up.
Velocity on its own is vanity. The honest measure is whether the new engine made money. Revenue hovered in the $10.5M–$11.5M range through early-to-mid 2023 while the loop was being built and tuned, then climbed steadily from mid-2023 forward, peaking near $16M in January 2024 before easing slightly in February.
"AI doesn't replace the marketer: it removes the part of the job that scales worst, so the marketer can spend their time on the part that compounds."
The ThroughlineEvery Engagement
Every engagement starts the same way: a workflow audit, a time-capture, and a short list of pain points addressable with tools that already exist. Then we get into the room and rebuild the loop.