AWS migration of the developer platform

The S&P Global developer platform powered the workflows of 4,000+ engineers building on top of Market Intelligence data. The legacy stack was capping scale, and every cost conversation pointed at the same line item.

The platform was also the post-merger surface where 15 services and 50,000+ users collided. Toolchain fragmentation made every incident slower to triage and every onboarding harder than it should have been. Cost was the loudest pain, but reliability and developer experience were the ones quietly compounding.

A migration with this many internal customers can't be sold on cost alone. Developers need a guarantee that their integrations keep working. The first OCTO Dev Tools survey we ran made the pain concrete: a small number of workflow fixes drove an outsized share of escalations. That gave the migration a credible story beyond "cheaper AWS bill."

• Owned the product side of the AWS migration: scope, sequencing, success metrics, and developer experience guarantees.
• Co-designed the migration plan with engineering leadership so individual service moves didn't break the platform contract.
• Stood up ELK-based uptime monitoring and centralized logging to give us a real before/after baseline.
• Drove the change-management story so 4,000+ developers could move without rewriting their integrations.
• Standardized the post-merger DevOps stack across 15 services and 50K+ users so the migration didn't ship more fragmentation.
• Ran the first OCTO Dev Tools survey and used the findings as the prioritization lens. Workflow fixes cut incident escalations 60%.

• 20% reduction in operational cost across the developer platform.
• 60% reduction in incident escalations from cleaner observability and faster rollback.
• 25% improvement in delivery velocity across distributed teams.
• A migration completed without a major developer-facing outage.

Cost was the headline. The real durable win was the change-management posture: every decision was tested against "what will 4,000 developers feel on Monday morning." That framing is what kept escalations falling instead of spiking during the cutover.

When I now scope AI rollouts inside enterprises, the same instinct applies. The model isn't the product, the day-after-launch developer experience is.