Why I Left Google Maps to Fix Mortgages

I grew up in a family of real estate developers, which means I absorbed two things at the dinner table: an unreasonable amount of opinion about property, and a front-row view of how much of the industry runs on phone calls, PDFs, and waiting. The mortgage was always the slowest part. Not the decision — the process.

My career then took what looks, in hindsight, like a suspiciously deliberate path toward doing something about it. At Goldman Sachs I was an analyst in Rates Systematic Market Making, building pricing models and systematic algorithms for interest-rate markets — the exact math that determines what a mortgage costs. At Google I spent three years as a software engineer on Google Maps, shipping consumer products used by billions of people, and learning what it takes to make complicated infrastructure feel effortless.

A mortgage sits precisely at the intersection of those two jobs. It is an interest-rate product, priced off the same curves I modeled at Goldman, delivered through a consumer experience that would not have survived a single Google design review.

Can software really handle the weird loans?

Mortgage people love making fun of outsiders who try to automate mortgages. The joke is always some version of: “they clearly haven’t seen enough weird loans yet.” I understand the instinct. Every loan file really does contain surprises — a borrower with three income streams, a condo with a strange HOA, an appraisal that comes back sideways.

A large pile of rules, exceptions, and edge cases is not evidence that software can’t do a job. It is the job description of software.

Listen to what the joke actually claims: if you knew how many rules, exceptions, and edge cases there are, you would give up. That is a fascinating argument to make in 2026. Codified rules are the easiest thing in the world to automate; the industry has simply been encoding them in people’s heads and three-ring binders instead of in systems.

What we’re building at Ralo

That belief became Ralo — in two steps. We entered the industry in 2025 with Approval AI, a platform that shopped mortgage offers for borrowers, and took it through Y Combinator’s first Spring batch. Shopping alone couldn’t fix the process, so we got licensed — I’m a mortgage professional under NMLS #2695701, and yes, the outsider went and got the insider credential — became a brokerage, and relaunched as Ralo in June 2026 with a $2.9M seed round.

My co-founder Arjun and I run the brokerage with a team of two. Both founders. We’ve rebuilt most of the consumer mortgage experience and automated a large share of the actual work: the rate shopping, the document chasing, the status updates that traditionally require a “team” of people forwarding each other emails.

I left a very good job building products billions of people use because the mortgage is the largest financial transaction most people ever make, and it is still priced and processed like it’s 1996. The weird loans are real. They are also exactly why this should be software.