At daily.jobs I was the founding technical program manager, which meant building the program management practice at the same time as the platform. The system had to score 65 million candidates in under ten seconds and serve data across 50,000 clients. At that scale, architecture decisions and program decisions stop being separate things.
The first lesson of founding-stage TPM work is that there's no process to inherit. You're not optimizing an existing rhythm, you're deciding what the rhythm is. I kept it minimal on purpose. Just enough structure to keep engineering, product, and data moving together, and not an ounce more, because process you don't need is just drag on a small team.
The second lesson is that performance at scale is a sequencing problem as much as an engineering one. Scoring 65 million records fast wasn't one clever trick. It was a chain of decisions that had to land in the right order, and my job was holding that order while the team built.
I also ran the A/B experimentation that lifted engagement by 35 percent. Founding work teaches you what's essential, because you have to build it from nothing and you feel the cost of every extra step.