Stories
One story per principle: ambiguity navigated, a bet that paid off, a bet that didn't, the hardest technical decision I've owned, and the time I grew someone. Each piece is the version I'd tell in a behavioral loop — context, scope, the call I made, and the number that came out.
- Ambiguity navigated
When the rules had no owner
The hard part of a rating-engine re-platform isn't the engine — it's that nobody in the building can tell you what the rules are. The path forward isn't a spec. It's a chair next to the actuary.
- Bet that paid off
Caching for correctness, not speed
Every outside reviewer told me to scale the database. I bet the cache was load-bearing for the contract — not just the latency — and that's why it worked.
- Bet I almost made
The job-per-table bet I almost made
I spent two weeks defending the obvious design before I noticed I was the one defending it. The real failure wasn't the architecture — it was how long I waited to change my mind.
- Hardest tech decision
When the cache becomes the contract
Moving the source of truth from Aurora to DynamoDB for a regulated workflow is the kind of decision that looks reckless on a whiteboard. The wrong call would have shipped duplicate clearances against real customer money.
- Time you grew someone
The shift from triage to root cause
The fastest way to grow an operations engineer isn't a training plan. It's to stop rewarding the work that keeps them stuck.