Transforming a Global SaaS Platform
A monolithic, batch-oriented platform rebuilt as distributed services, with a data platform organization built alongside it and cost treated as part of the design.
Transformations
Seven pieces of work, spanning enterprise infrastructure, hybrid cloud, multi-cloud SaaS and the organizations built around them.
Each is written the same way: what the situation was, what made it difficult, what was decided, and what the decision cost. The costs are the part worth reading. Every choice here bought something and gave something up, and the ones that went well were not the ones that came without a bill.
They are ordered by what they show rather than by when they happened.
~80%lower cloud infrastructure cost
Transforming an Enterprise Platform
An acquired product re-architected for horizontal scale while enterprise customers continued running it in production.
Thousands of hosts. Hundreds of thousands of virtual machines.
Engineering for Economics
After the strategy changed, the engineering problem became delivering the same customer value while the platform cost materially less to run.
First Engineer to Organization
A new site with no local knowledge of the product it had been given, and an open question about whether it could be trusted with it.
1 → 4 → 10 → ~100
Building Hybrid Cloud Across Boundaries
A managed-database experience extended into on-premises environments, architected jointly by two companies where neither controlled the whole system.
~1 yearfrom concept to launch
Turning Platform Integration into Product Adoption
A capability with a small install base, sitting beside a workflow that a far larger customer base already used every day.
Innovation From the Engineering Floor
Two prototypes built outside the roadmap, for problems nobody had assigned.
Technology
Not a list of everything I have touched. These are the technologies the work above actually ran on, grouped by the problem they were solving.
Distributed systems
Java, Spring Boot, Cassandra, GemFire, PostgreSQL, Lucene — the foundation of the enterprise platform rebuild, where three data technologies were chosen because three workloads wanted different things.
Cloud
AWS, Azure, Google Cloud, OCI, VMware vSphere — normalizing cost and usage data that each provider describes differently, and extending a managed service into an environment its designers did not control.
Data and analytics
Distributed SQL, event-driven pipelines, data lineage, multi-cloud normalization — the spine of the SaaS modernization, and later the thing whose economics had to be unwound.
Platform efficiency
Work avoidance, caching, workload routing, compute-efficient instances, retention policy — mechanisms rather than a directive, when the constraint became margin instead of capability.
AI
Conversational analytics grounded against the platform’s known data and schema before questions become executable queries, and operational intelligence inside engineering operations.
Technology is a means to an outcome, not the outcome itself.