4 agents working one pipeline, live

Four agents that turn public data into deals.

Scout, Underwrite, Outreach and Structure run as one autonomous pipeline across distressed properties and land — from public records to a Buy / Pass / Watch brief on every parcel.

Book a demo See the pipeline
13-rule distress engine XGBoost 0–100% scoring LLM underwriting briefs
0
AI agents in the pipeline
0
Distress rules, two tiers
0
ML features scored by XGBoost
0
Parallel data providers
The pipeline, agent by agent

One deal, four agents, no dead ends.

Each agent owns a stage and hands off to the next — a parcel enters as a public record and leaves as a scored, underwritten, structured deal.

STAGE 01

Scout

Sweeps three parallel data providers to surface distressed properties and land parcels the market hasn't priced yet.

  • ATTOM, RESO MLS & probate signals
  • SFR, multifamily, commercial & land
STAGE 02

Underwrite

Runs the 13-rule distress engine and XGBoost scoring, then an LLM agent writes a Buy / Pass / Watch brief you can read in seconds.

  • 0–100% distress score, 11 features
  • Explicit rules plus ML, side by side
STAGE 03

Outreach

Turns a qualified parcel into contact — an LLM agent drafts owner outreach off the underwriting brief so scored deals don't stall in a spreadsheet.

  • Grounded in each property's brief
  • Prioritised by distress signal
STAGE 04

Structure

An LLM deal-structuring agent picks from a 9-strategy library to shape the offer that best fits the property, the owner and the numbers.

  • 9-strategy deal library
  • Matched to asset class & brief
How distress is scored

Explicit rules you can read. ML that ranks them.

Every property is judged two ways — a transparent 13-rule classifier and an XGBoost model over 11 features — so a Buy verdict is never a black box.

13-rule distress engine Rule-based

Explicit distress classification across two tiers — the signals that flag a property before a model ever sees it.

Tier 1 · Critical distress signals
01 Pre-foreclosure 02 Tax delinquency 03 Active liens 04 Probate / estate 05 Vacancy 06 Code violations
Tier 2 · Supporting distress signals
07 Absentee owner 08 Long ownership 09 High equity 10 Below-market value 11 Failed listing 12 Ownership change 13 Condition flag

XGBoost scoring 11 features

An XGBoost model served on Flask turns the signals into a single 0–100% distress score that ranks the whole pipeline.

Owner & occupancy signals0.88
Financial & equity position0.74
Legal & tax status0.63
Property condition & age0.52
Market & listing history0.41
Buy The LLM agent writes the brief from these features — Buy, Pass or Watch.
Inside the product

The pipeline, on one screen.

Every scored parcel, every brief, every map pin — the work of four agents laid out for the analyst who signs off the deal.

THE PIPELINE VIEW

Scout, underwrite, outreach and structure — in one board.

The dashboard is the fund's control room: parcels flow left to right through the four agents, each carrying its distress score and current stage.

  • Live distress scores across the pipeline
  • Stage-by-stage status for every deal
  • Filter by asset class and subscription tier
The deal-pipeline dashboard showing distressed properties moving through the scout, underwrite, outreach and structure stages with distress scores.
The pipeline board — scout, underwrite, outreach and structure, each parcel scored.
THE PROPERTY BRIEF

Why this parcel is distressed — and what to do about it.

Open any property to see the distress drivers that fired, its 0–100% score and the LLM-written underwriting brief with a Buy, Pass or Watch verdict.

  • The distress drivers behind the score
  • LLM underwriting brief with a clear verdict
  • 3D Mapbox context with a Street View popup
A single property detail view showing its distress drivers, the 0 to 100 percent distress score and the LLM-generated underwriting brief.
A property detail — distress drivers, its score and the underwriting brief.
Data in, deals out

Everything the fund runs on.

Three data feeds, a scoring stack, a mapping layer and a strategy library — gated by subscription tier across every asset class.

Three parallel data feeds

ATTOM across 8 endpoints, RESO MLS through an OData client, and dedicated probate signal detection — running in parallel so the Scout never waits on one source.

ATTOM · RESO MLS · Probate

13-rule distress classification

An explicit, two-tier rule engine flags critical and supporting distress signals — the transparent layer beneath the ML score.

Two tiers · Explainable

XGBoost on Flask

A gradient-boosted model over 11 features, served on Flask, scores every property 0–100% and ranks the pipeline for the underwriting agent.

11 features · 0–100%

3D Mapbox & Street View

Every parcel lives on a 3D Mapbox map with Street View popups, so the analyst sees the block before the offer, not after.

3D map · Street View

9-strategy deal structuring

The structuring agent draws from a 9-strategy library to shape each offer around the property, the owner and the underwriting brief.

LLM agent · 9 strategies

Tier gating by asset class

Subscription tiers gate access across SFR, multifamily, commercial and land — each user sees the asset classes their plan covers.

SFR · Multi · Commercial · Land
Built on
Python LLM agents XGBoost Mapbox Flask ATTOM API
Talk to us

See four agents run a real pipeline.

We'll walk you through Scout to Structure on live distressed inventory — the distress engine, the scores and the briefs, end to end.

Book a demo Talk to the team