Framework — Proprietary AI Underwriting Language
The Engine, Decompiled

Five passes.
One fundable file.

Every deal moves through the same five passes, in the same order. Each pass takes structured input, applies one underwriting discipline, and produces output the next pass can consume. By pass five, the file speaks the language capital sources actually underwrite in.

pipeline.flow
# five passes, deterministic order
01 parse → structured inputs
02 score → sponsor grade
03 model → risk index
04 recommend → ranked fixes
05 package → lender-ready
runtime 3m 48s
5
Sequential Passes
47+
Signals Evaluated
12
Output Fields
< 4m
Median Runtime
PASS 01

Parse the deal.

Most files arrive as a phone call, a sloppy memo, or a screenshot of an LOI. Pass one strips the narrative and resolves the file into the structured inputs every downstream pass requires — asset, sponsor, leverage, timeline, exit.

// what it normalizes

  • Asset class, location, current state (stabilized, value-add, ground-up)
  • Sponsor identity, entity structure, signing authority
  • Capital stack: requested debt, equity in place, mezz
  • Timeline: close-by date, hold period, exit milestone
  • Disclosed risks: prior bankruptcies, environmental, title
01_parse.io
# raw input
"$12.5M multi, 80 units,
Tampa, value-add reno,
sponsor has 3 prior."

# parsed output
asset.class multifamily
asset.units 80
asset.state value-add
asset.market Tampa FL
sponsor.prior 3 verified
debt.req $12,500,000
PASS 02

Score the sponsor.

The first signal every lender reads is who the borrower is. Pass two resolves experience, liquidity, and track record into the Sponsor Strength Rating™ — the A–F grade that determines which capital sources the file is even worth showing.

// what feeds the grade

  • Verified prior projects in the same asset class & market
  • Post-close liquidity (cash plus liquid securities)
  • FICO & credit events within the last 36 months
  • Net worth relative to deal size
  • Operating partner depth (property mgr, GC, sponsor team)
02_score.io
# sponsor inputs
experience 3 prior, on-thesis
liquidity $2.1M (16.8%)
fico 744
net_worth $8.6M

# grade output
sponsor_strength A−
tier_access institutional
flags none
PASS 03

Model the risk.

Pass three forecasts whether the asset reaches stable, lender-accepted cash flow on the projected timeline. The Stabilization Risk Score™ is the index lenders use to decide reserves, interest-only periods, and exit ratios.

// what's modeled

  • Lease-up velocity vs. market comps over a rolling 24 months
  • Cap-ex burn schedule and contingency adequacy
  • Rent roll fragility (concentration, lease maturities)
  • Market trajectory: supply pipeline, rent growth, vacancy
  • Macro overlay: rate environment, refi market depth
03_model.io
# risk model
leaseup_months 14 (mkt: 11–16)
capex_burn $1.8M / 18 mo
contingency 8.2%
refi_window Q3 ‘27

# risk output
stabilization_risk 28 / low
reserve_rec 12 mo IO
PASS 04

Structure the recommendation.

Pass four is the difference between a 78 and an 86. The engine returns the specific, ranked adjustments — leverage, reserves, sponsor cures, structural moves — that lift the file's grade before it goes to capital.

// what gets recommended

  • Leverage adjustments: LTV/LTC ceilings that unlock better-priced debt
  • Reserve sizing: interest reserves, opex reserves, capex pools
  • Sponsor cures: KP additions, liquidity top-ups, signer changes
  • Structural moves: mezz vs. preferred, IO period, recourse carve-outs
  • Documentation gaps lenders will demand before LOI
04_recommend.io
# weak points detected
issue.01 LTV at 75% blocks tier-1
issue.02 no interest reserve

# ranked fixes
fix.01 drop LTV to 68%
impact +7 grade pts
fix.02 12mo IO reserve
impact +4 grade pts
projected_grade B+ → A−
PASS 05

Package for capital.

The final pass is the artifact. A single document that opens with the scored signals, walks through the structure, surfaces the risks the engine already identified, and closes with the lender list pre-filtered to capital sources that fund this exact profile.

// what ships

  • One-page scored summary — the page lenders read first
  • Sources & uses table with the recommended structure applied
  • Sponsor sheet with verified track record & references
  • Risk register: every flag the engine raised, with mitigants
  • Capital fit list: ranked lenders/investors matched to this profile
05_package.out
# deliverable assembled
deal.summary 1pg / scored
sources_uses structured
sponsor.sheet verified
risk.register 7 flags
capital_fit 14 lenders

grade A−
funding_probability 82/100
status ready to submit
Translate Your Next Deal

Put your file in the language lenders fund.

Submit a scenario and see it scored, structured, and packaged the way capital sources actually underwrite — in minutes.