Andrew Wang

Most founders chase the future. Andrew Wang went after the past - the 1960s mainframe logic still humming underneath your monthly mortgage statement. The mortgage servicer is the company that takes your payment, holds your escrow, and is supposed to talk to you when life goes sideways. For decades it has been the least-loved corner of finance: commoditized, ignored, and running on code older than most of the people maintaining it.

Wang's diagnosis is unsentimental. "The business is pretty commoditized," he has said. "It's what ends up happening when you have a very old school faced with little innovation." His fix was not a friendlier app bolted onto the old machine. It was the machine itself. Valon, the company he co-founded in 2019, threw out the wrappers and built its own system of record from scratch - ValonOS - and then went and got it licensed by Fannie Mae. Valon was the first new servicer to do that on a proprietary system, a regulatory milestone roughly equivalent to showing up to a vintage-car rally with a car you welded yourself and being handed a racing license.

The path there reads like a deliberate accumulation of unfair advantages. Wang earned a BA and a master's in Computer Science from Harvard. He interned at the Executive Office of the President - the White House - in 2011, and as a software engineer at Google in 2012. He started his finance career at Goldman Sachs in, of all places, the aviation finance group, learning to underwrite airplanes before he ever underwrote a home. Then came Soros Fund Management, where he rose to principal investing at the intersection of private credit and venture, staring directly at mortgage assets as an asset class. He had seen the industry from the capital side. He knew exactly where the money leaked.

The bet: engineering, not interest rates

Wang describes Valon as a "high alpha, low beta" business. Translated out of hedge-fund dialect: the company is not betting on which way mortgage rates move. It is betting that software can drive the cost of servicing a loan down far enough to win regardless of the macro weather. That is a quiet, almost stubborn thesis. It means the upside comes from craft - from correctness and automation - rather than from catching a wave.

He is equally precise about how startups actually move. His rule of thumb borders on heresy for perfectionists: making 15 decisions at 75% accuracy beats making 5 decisions at 90%. The point is velocity. "Growth solves all problems in startups," he says, and he means it as both a strategy and a discipline - keep putting shots on goal, keep shipping, and let momentum dissolve the problems that paralysis would only calcify.

That philosophy shows up in the numbers. Valon has posted 400% year-over-year growth, crossed $65 billion in mortgages serviced on its platform, and - in a detail that makes traditional lenders wince - reportedly sold its first $100 million of business without a dedicated sales team. The product was the pitch. When you service roughly 1% of all American homeowners, word travels.

The spaghetti-code problem

Wang has a name for what makes regulated industries so resistant to change: the spaghetti-code problem. Mortgage, healthcare, banking - each runs on systems 30 or 40 years old, then patches them with hundreds of integration "wrappers" until the whole tangle is too brittle to touch. Most startups respond by adding one more wrapper. Valon's contrarian move was to refuse the shortcut and build the underlying system of record itself. a16z general partner Angela Strange put it plainly: the founders "took the much harder, durable path."

That durability extends to how Valon thinks about artificial intelligence, which is everywhere in its stack - Anthropic's Claude and a modern data spine of dbt, Airflow, ClickHouse and Vitess sit alongside the servicing engine. But Wang applies a filter he calls the "Congress test": before AI touches a regulated decision, a regulator has to be able to understand why the software decided what it did. Large language models hallucinate and behave non-deterministically, which is charming in a chatbot and disqualifying in a foreclosure workflow. So at Valon, AI accelerates the work it can explain, and stays out of the decisions it cannot.

What he is building toward

Mortgage servicing is the wedge, not the destination. The longer ambition is to make ValonOS the default operating system for regulated finance - the same correctness-first, API-driven foundation extended to other industries strangled by their own legacy. The October 2024 Series C, $100 million led by WestCap with Andreessen Horowitz, was explicitly raised to accelerate platform development and market expansion. WestCap's Josh Dart took a board seat. The capital is fuel for a thesis that has not wavered since 2019: that the most overlooked plumbing in the economy is exactly where a software company should plant its flag.

There is something fitting about an engineer who once financed airplanes and once worked inside the White House deciding that the frontier worth claiming is the mortgage statement. It is not glamorous. It does not trend. It is, by Wang's own description, old-school and commoditized. Which is precisely why he wanted it. "What needs to happen," he says, "is actually a large leap in technology." He is not waiting for the industry to take it. He is building the runway, the plane, and asking the regulator for the license - all at once.