Staff Software Engineer, Ledger

Alt is unlocking the value of alternative assets, starting with the $5 B trading-card market. We let collectors buy, sell, vault, and finance their cards in one place and we are backed by leaders at Stripe, Coinbase, Seven Seven Six, and pro athletes like Tom Brady and Giannis Antetokounmpo. Our next frontier is real-time pricing at scale—the Alt Value that powers every trade, loan, and product on the platform.

THE NORTH STAR

99.99% reliability on gateway and banking integrations, and a ledger where every cent is accounted for and every entry says who created it and when.

Supporting KPIs: reconciliation exception queue depth at daily close, gateway and banking integration uptime at 99.99%, P90 time from auction close to seller funds available, share of ACH returns and reversals handled with no manual correcting entry, and time to bring a new PSP or rail to production.

TH ROLE

This is a builder's seat, not a manager's seat. You will write the ledger, the rails, and the reconciliation yourself. You will make the build-versus-buy call and defend it. You will tell the business no when correctness demands it, and you will be right often enough that they keep asking.

The people who are great at this have owned a ledger somewhere it mattered, and they think in idempotency, race conditions, and reconciliation.

WHAT YOU'LL OWN

• The ledger as system of record. Entries, lines, balances, holds, reversals, and the states between: pending, posted, available, canceled, declined, reversed. Immutability and audit trail as design constraints, not features. Point-in-time balances that show a customer where they stood on any past date. Done looks like an auditor pulling a trial balance without asking you for help.

• Rails and money movement. ACH origination and returns, RTP and other instant rails, wires, cards, and stablecoin settlement. A credit-push rail changes what you can auto-charge and when funds are final, and you own those product consequences, not just the integration. Done looks like the fourth PSP costing a sprint, not a rewrite.

• Reconciliation. When the ledger and the processor disagree, you own the answer. Automated matching, an exception queue a human can actually work, and breaks caught before a customer finds them. Every point of automation here is a point on the north star.

• Payouts and lending infrastructure. Getting sellers and borrowers their money faster than anyone else in this market: instant payout options, batch payouts at auction close, and the settlement timing that makes them safe. Then the full loan lifecycle, disbursement through repayment, partial payment, delinquency, and collateral release.

• Risk posture on high-value flows. Holds, limits, and settlement windows set with data, and you will know what each one costs in conversion. Step-up verification triggered by risk signals rather than blanket friction.

• The trade-off calls. Shipping speed against uncompromising correctness, build against buy. You set the planning cadence with Robbe, you work across Finance, Risk, Ops, and Support, and you are the one who explains upward why a number moved.

WHAT GREAT LOOKS LIKE IN 6 MONTHS

• The reconciliation automation rate is materially higher than the day you started, and you can show the weekly series.

• The exception queue at daily close is worked to zero by a human in under an hour, on a system you built for that.

• A new rail or PSP has gone to production on the provider architecture you designed, and it did not require a ledger migration.

• Pending, posted, and available are modeled cleanly enough that Support answers "where is my money" off the customer-facing surface instead of escalating to engineering.

• Corrections post as reversing entries. Nobody has edited history, and an auditor can say so.

• At least one thing you shipped is running with AI in the loop, and the team depends on it.

• Finance trusts the close.

HOW YOU'LL USE AI HERE

• Turn a payment provider or bank partner integration spec into a typed client and a test suite in an afternoon instead of a sprint, including the ACH return codes and reversal paths nobody wants to hand-write.

• Stand up a rail simulator that produces the failures a sandbox won't: a late return, a duplicate callback, a partial reversal, a webhook that arrives twice out of order, a settlement that lands after the window closed. Then run your ledger against it.

• Build an agent that triages the reconciliation exception queue overnight, clusters the breaks, and hands you a ranked list with a proposed cause for each instead of a wall of unmatched rows.

• Put an MCP server over the ledger so "where did this discrepancy co