Moonxi
Book a diagnostic
← Case studies Case · Delfin One · Construction credit

A construction credit platform, with 44 AI agents on documents, risk and market.

A continuous engagement, not a closed project. Moonxi builds it and runs it.

Context

Construction credit: a mistake is expensive on both sides.

Every application arrives as a stack of documents: incorporation papers, the construction budget, the schedule, the collateral. The analysis has to read all of it, price the risk of default and of cost overrun, and answer fast. Read slowly and you lose the client; read wrong and you lose the money.

What was built

Moonxi built the whole platform, and runs it.

44 analysis agents

17 sit in the credit path — contract, budget, schedule, collateral — and every reading carries its source. The other 27 handle market analysis and chart generation.

A proprietary real-estate credit model

Built for this business, with the calculation deterministic and tested. The model writes the text around the number, never the number.

A machine-learning engine

Probability of default and of cost overrun, compared against the deterministic path before any of it reached production.

Pipelines and backend on AWS

With infrastructure as code: the entire environment can be rebuilt from the repository.

Architecture

From application to decision.

input Application and documents
agents Document and risk reading, with the source attached
engine Credit model + ML for default and overrun
approval A person approves before anything goes out
decision Auditable line by line
What runs today

A continuous engagement, not a closed project.

Moonxi still operates the platform: watching the agents, tuning the models and answering for what is live. Every number carries its source, anything that is a calculation stays deterministic, and a person approves before an action goes out.

Want this inside your operation?

It starts with a diagnostic: two weeks, one document, one fixed quote.

Book a diagnostic →