Founding Security Engineer

Casca is building AGI for banking. We’re replacing decades-old legacy systems with AI-native technology that automates 90% of the manual work humans once had to do.

Our customers are FDIC-insured banks. Every one of them puts us through security review, control questionnaires, pen test results, and third-party risk assessment before we touch production. Then they hand us borrower tax returns, bank statements, and the data behind credit decisions on their own balance sheet.

So security isn't a compliance checkbox here. It's a precondition for the company existing. Every deal we win depends on it, and one bad breach ends the business.

The surface area is large and getting larger: a multi-tenant AI platform, a cloud estate under bank data-handling requirements, agents that read and act on customer data, and an engineering team shipping into all of it every day. Nobody can review their way across that. You build the guardrails, defaults, and tooling that scale past you.

Software. Detections, libraries, policy-as-code, automation, CI/CD checks — reviewed, tested, and deployed like any other service.

Design reviews and threat models on what actually matters: auth, tenant isolation, data handling, anything touching borrower PII.

The AI attack surface. Prompt injection, tool-use abuse, exfiltration paths through agents, and the evals that tell you whether your controls hold.

Incidents, end to end, including the write-up and the fixes that follow.

Bank security reviews and audits, turned into engineering work that improves the system rather than paperwork that describes it.