Sunbit brought me in to help their data science team get better at software engineering practices, and at collaborating more tightly with the backend team. Earning that trust is what let me, later, propose a real-time platform for tracking purchases, based on technologies suited to that domain that they hadn’t used before: Event Sourcing on Kafka, with Kotlin and Kafka Streams.

The technology was hard. Kafka Streams (as opposed to Kafka itself) was immature, and I pushed relentlessly to make it do what we needed. I succeeded in doing so, and then got a lot of pushback from the backend team when I tried to get them to adopt what I’d built.

That was the part I got wrong. I had proved the technology could work, and skipped the harder question of whether this team could support it. They were right: they couldn’t yet own something that fragile.

Years later, I heard they adopted it, once the technology had matured enough for them to run it themselves.