For two decades I watched institutions pay handsomely for broad market visibility while the questions that actually move credit decisions went unanswered. Who controls the asset. Who holds the debt. Where refinancing pressure is building. How one sponsor’s exposure connects across the capital stack. The records existed — public, filed, available — and no one had assembled them into anything you could use.
Alpine Systems is what I built to answer those questions.
Two decades inside institutional CRE and finance, including venture-backed and Tier-1-funded businesses built where software, data, and investment decisions were meant to work together. In practice they rarely did. The recurring frustration was always the same: the intelligence that mattered most stayed fragmented across records nobody troubled to connect.
That experience isn’t incidental to what Alpine does — it’s the input that makes it work. The rules governing how CRE debt is actually reported aren’t written in any manual. Statutory filing schemas, disclosure-regime quirks, sponsor-structure conventions, amortization mechanics, capital-stack taxonomies — these are tacit knowledge, learned by sitting inside the market long enough to see how filings really behave versus how they’re supposed to.
Knowing where a filing misleads, which field hides the real number, how an SPV is typically nested beneath its sponsor — that judgment is what separates a pile of parsed records from intelligence you can underwrite.
Alpine is the product of eighteen months developing and testing the extraction logic, the normalization approach, and the resolution methodology — combining that domain knowledge with serious data engineering and modern AI capability.
The division of labor is deliberate. The data and AI tooling handle scale: reading filings, structuring records, holding the graph together, surfacing what’s changed. Domain judgment handles correctness: deciding what a record actually means, where a structure is ambiguous, when a link earns a high confidence score and when it doesn’t. The tooling does the volume. The expertise governs the truth.
That balance is the entire point. Tooling alone produces confident nonsense at scale. Expertise alone can’t reach across trillions of dollars of debt. Held together and refined over eighteen months against real filings, they produce something neither could alone.
How we build it →Everything traces to a source. Every relationship carries a confidence level. Nothing enters the system that can’t be tied back to a filing. Not as a feature to advertise — but because intelligence an institution is going to act on has to be defensible to the person acting on it. That standard governed every decision in the build, and it’s the one I’d want if I were on the other side of the screen.
Modern tooling is becoming widely available. That’s a good thing, and it’s the reason a build like this is possible now. But the tooling was never the hard part. The hard part is the domain knowledge that decides what the data means — and that doesn’t commoditize. It compounds.
The longer the platform runs, the deeper the graph, the tighter the confidence, the fewer the gaps. An advantage built this way widens quietly.
Strategic partnerships and operational access run on different commercial terms — both start with a conversation with me directly.
Email Andrew
