Skip to content
Tasklet

Mālama Funding · US business-purpose lender for real estate investors

Loan sizing and property intelligence inside a lending platform

A loan origination platform for investor loans, where we engineered the loan sizer, loan-product configuration and property search that staff use to price and qualify deals.

Backend engineering contribution
Mālama Funding
Mālama Funding staff dashboard with loans needing attention, tasks and calls

What we did, plainly

Mālama's platform was built by a larger team. Our co-founder Birehan Anteneh contributed backend engineering between March and August 2025. Items tagged “Our work” below are the parts we engineered. Everything else is shown so you can see the whole system, not because we built it.

157
application screens across staff and borrower portals
640+
API operations in the backend
5
loan strategies modelled
4
background workers for sizing and document AI

These figures describe the whole platform, which was built by a larger team. Our contribution is the backend work tagged “Our work”.

The problem

Investor loans (fix and flip, bridge, rental, new construction) are priced against many variables: the property, the borrower, the loan product and the lender's rules. Teams doing this in spreadsheets and email re-key the same data and answer the same questions slowly.

A lender also has to track the file itself: contacts, documents, checklists, communications and what the borrower has been told.

Why off-the-shelf was not enough

Generic CRMs do not model loan products, pricing assumptions or the property diligence that drives a decision. The platform needed a configurable engine for products and fields, not a fixed form.

What the platform does

The platform gives staff a loan workspace, a borrower portal and a set of background workers. The loan sizer produces loan options from a deal's inputs. Loan products and their fields are configurable, so new products do not need a developer.

Property search with sold comparables and due-diligence data helps staff judge a deal before a term sheet goes out.

How it works

Inside the platform, step by step

  1. 01 / Staff dashboard

    Start the day from what needs attention

    Loans needing attention, the day’s tasks and a ranked call list sit on one screen, so staff open the platform and know what to do next. Names in this screenshot are blurred.

    Live
    Mālama Funding
    Staff dashboard with loans needing attention, tasks and calls, names blurred
  2. 02 / Property search

    Judge the property before the term sheet

    Property search with nearby sold comparables and due-diligence data puts the evidence behind a loan decision next to the file.

    LiveOur work
    Mālama Funding
    Property search with comparables
  3. 03 / Configurable products

    New loan products without a developer

    Loan products, their fields and validation rules are data, not code. Dynamic custom fields let staff capture what a product needs.

    LiveOur work
    Mālama Funding
    Loan product configuration screen
  4. 04 / Documents and checks

    Every document filed and tracked

    New files land in an inbox and are filed against a document type. Requests for documents the borrower still owes sit beside what has arrived. Background workers index documents and run AI checks against checklist items.

    Live
    Mālama Funding
    Loan document inbox asking what a newly arrived document is, with requests and review lists

Screens are from an environment populated with test data. No real customer, agent or borrower records are shown.

Capabilities

What exists, and what does not.

The whole platform is listed so you can see what a lender can have. Items marked “Our work” are the parts we engineered.

Live
Running in the client's production environment.
Built
Implemented in the codebase, but switched off for production users or waiting on configuration.
Could build
Not built. Something we could build for you, shown so you can see where the work could go.
  • Loan sizer and loan options

    Inputs to options, with an automatic run on new inquiries.

    LiveOur work
  • Loan product configuration and dynamic fields

    Products, fields and validation as data.

    LiveOur work
  • Property search, comparables and due diligence

    Evidence beside the file.

    LiveOur work
  • Document history and requests

    What was asked for and what has arrived.

    LiveOur work
  • Staff loan workspace and pipeline

    Files by stage with tasks and notes.

    Live
  • CRM contacts, email and messaging templates

    Communication tied to the loan.

    Live
  • Borrower portal and term sheet links

    Borrower-facing views and shareable term sheets.

    Live
  • AI document indexing and checklist validation

    Background workers on Google Cloud Pub/Sub.

    Live
  • Construction draw management

    Not part of the platform. A realistic extension for construction lenders.

    Could build
  • E-signature on loan documents

    Not built. Integration with an e-sign provider would be an extension.

    Could build

Operational value

What it changes day to day.

Described from how the system works. We do not quote savings or percentages we cannot evidence.

Pricing logic that lives in the system
Assumptions and loan products are configured once and reused across every deal, instead of living in a spreadsheet.
Faster first answers
Automatic sizing on a new inquiry means staff start from loan options, not a blank form.
Evidence next to the decision
Comparables and diligence data sit in the loan workspace rather than in browser tabs.
Products that change without releases
Configurable products and fields mean the lender can launch or adjust a product on its own timetable.

Built with

  • Next.js and TypeScript
  • FastAPI on Python
  • PostgreSQL with pgvector
  • Google Cloud Run and Pub/Sub workers
  • OpenAI models via LlamaIndex

Next

Al-Shalawi Law

Saudi law firm · bilingual legal operations

Read the Al-Shalawi Law case study

Pricing and qualifying deals in spreadsheets?

Tell us what your team prices, checks and chases by hand. We will tell you where software or automation would help, and where it would not.