Augustin Delaporte

Problem. Platform.sh was engineering-led, and for a while that worked: the founders and the engineers closest to the code made the calls. Growth broke it. New segments — enterprise, e-commerce, agencies — opened faster than anyone could own them, and there was no product function to hand prioritisation, discovery or the product-market-fit conversation to.

Approach. Moving the company from engineering-driven to product-driven, while I was learning the role in the middle of it. We hired the first PMs, wrote the operating model as we went — discovery cadence, RFC process, quarterly planning — and grew the team. We got the structure wrong several times: ownership boundaries, reporting lines, how product paired with engineering. Each version held until the company outgrew it, and then we changed it again instead of defending it.

Result. 20+ PMs and Directors across three continents, and product-driven decisions at a company that had been engineering-driven for its first years. That organisation went on to absorb a brand merger (Upsun), a shift to product-led growth, an acquisition (Blackfire) and a new product line (Upsun Dispatch). The company grew 4 → 350+ people and €0 → €50M ARR over the same period. Change was the only constant through those years, and a growth mindset across the team is what turned it into an advantage rather than churn.

Problem. Developers increasingly work through AI assistants and IDEs, but the platform's interface assumed a human at a CLI or console. Without a machine-usable surface, Upsun would be invisible to the fastest-growing part of the developer workflow.

Approach. Framed "AI-native" as three concrete bets: (1) AI-assisted configuration, so intent — not YAML — is the input; (2) an MCP server exposing platform operations to any assistant; (3) agent-based troubleshooting for common failure modes. Each bet was prototyped hands-on to pressure-test feasibility before staffing the work.

Result. The Upsun MCP server shipped, AI configuration APIs reached the CLI and IDE extensions, and agent-driven workflows went from prototype to delivered capability — automating configuration, troubleshooting and operational tasks.

Problem. Customers and partners — mostly e-commerce — kept asking us to open a cloud region in mainland China. Serving that demand was never just an infrastructure question: it required a local legal entity, a team on the ground, and cloud operator licences, none of which are formalities there.

Approach. I went out because I spoke Mandarin and had worked in China earlier in my career. We set up the entity, ran the licensing process, and worked the customer and partner channel directly rather than through the standard go-to-market motion. The architecture and the stacks had to change too: development environments secured by default to meet local regulation, and proxies inside the Great Firewall so builds could still resolve their dependencies.

Result. Enterprise customers ran in production, so the demand was real. The licences were not: we never obtained the cloud operator permits the region depended on. In parallel our partners' priorities moved — Adobe's acquisition of Magento redirected our largest e-commerce channel. After two years we made the call to stop, and closed the legal entity. The build was sound and the demand was there; the regulatory path wasn't, and reading that honestly mattered more than keeping the project alive.

Problem. Magento wanted a managed cloud offering and had failed to ship one more than once — in-house, and with third-party partners. Platform.sh had the technology but nothing that could carry someone else's brand: no white-label product, no partner-scale support model, and gaps in enterprise compliance.

Approach. Treated it as an engineering problem rather than a branding exercise. I spent two months embedded in Magento's offices in Austin, working directly with their teams, and we reworked several parts of our own architecture to fit their model — isolation, environment topology, support tiers and SLAs sized for a partner rather than an end customer. Where the stack needed capability we didn't own, we partnered instead of building — Fastly at the edge, Blackfire for performance profiling — so the result read as one product rather than an assembly. Compliance (SOC 2, HIPAA, PCI) was prioritised as a sales unlock, not a checkbox. Requirements and escalations flowed through a partner, which changed how the roadmap had to be run.

Result. Magento Cloud shipped and grew past 5,000 enterprise customers, and was part of the business Adobe acquired in 2018 for $1.68 billion — becoming Adobe Commerce Cloud. What started as one partner became a repeatable line of business: Symfony Cloud, Ibexa Cloud, Pimcore Cloud, Shopware Cloud and OVHcloud Web PaaS followed on the same foundations, and the compliance work became the basis for Platform.sh's own enterprise tier. Blackfire, first brought in as a partner here, was later acquired outright and folded into the platform.