The Settlement Season.

PayZoll began as a thesis on friction. Capital moves slowly between borders and blockchains. I was on it from the first prototype: through the Educhain Open Campus incubation, then the Stellar Build Better hackathon, then the build that carried the SCF application. A $100K grant from the Stellar Development Foundation turned that prototype into a company, and I stayed on the Stellar side of the engineering as a co-founder.

Engineering as a thesis.

I was the sole technical owner of the payment infrastructure, and the work was less glamorous than the pitch. Redesigning the Stellar Disbursement Platform from single-tenant into a multi-tenant distributed system. Moving signing keys out of raw token storage and into hardware-derived material in AWS KMS, so no organization's funds or data were ever exposed to another's. Rebuilding the transaction queue with distributed workers and a real-time event layer, because the old "optimistic done" status marked payments successful before on-chain confirmation, and a payment that lies about itself is worse than one that fails loudly. Then a fund-recovery service to unwind partial failures without losing funds.

The ceiling I remember best was arithmetic. The platform could hold eight channel accounts, which meant about two transactions per second, which meant a large payroll took longer than a payroll is allowed to take. Soroban's authorization-tree model turned out to allow one channel account to fan out into many parallel contract calls, and that moved it to roughly 200 per second. Rust and Go, learned as needed. The part that stayed with me is not any deploy, it is what a payment system has to get right at 3am on a Sunday, when there is nobody to reverse anything.