Upload a prospectus, tape, or VDR. Graam extracts the deal structure, builds an executable model, runs scenarios, and produces investment and legal outputs.

Inputs
User Query
Prospectus / Tape / VDR
Agent Layer
Extraction & QA

Document parsing, semantic search & citation-linked answers

Analysis & Modeling

Historical data, loss modeling & scenario projections

Structuring & Rating

Capital stack sizing, S&P criteria & waterfall validation

Flows Engine

Waterfall projection & bond analytics

Outputs
IC Report
Deal DSL
Legal Docs
Scenarios
Analytics

Phase 1 — Inputs & Extraction

Document Intelligence

Query any prospectus with inline citations

Ask questions about deal terms, waterfall mechanics, or risk factors. Graam retrieves the relevant passages and returns answers with page-level citations.

Semantic Search over Uploaded PDFs

Upload prospectuses, offering memoranda, or servicing reports and search across them with natural language queries. Graam indexes documents on upload and retrieves the most relevant passages.

Citation-Linked Answers

Every answer includes page-level citations pointing back to the source document. Click through to the exact passage that supports the response.

Multi-Document Queries

Query across multiple documents in a single conversation. Compare terms across deals, identify differences in waterfall mechanics, or cross-reference risk factors across prospectuses.

Term Search Across Deal Documents

Search for specific terms, definitions, or clauses across all uploaded deal documents. Find how a particular trigger is defined across multiple deals in your portfolio.

Tape Stratification

Pool composition at a glance

Upload a loan tape and get instant stratification across FICO, LTV, rate, term, geography, and asset type. Weighted averages, concentration metrics, and risk flags.

Stratification Tables

Stratifies across FICO, LTV, rate, term, geography, and asset type. Computes WFICO, WAC, WALT, and balance-weighted concentration metrics. Identifies risk concentrations that drive structuring decisions.

Weighted Averages & Concentrations

Computes weighted-average FICO, coupon, LTV, original term, and remaining term. Flags geographic and issuer concentrations that exceed typical thresholds and may require additional credit enhancement.

Risk Flags

Automatically identifies outlier loans, missing data fields, and concentration risks. Surfaces issues that affect pool classification and stress assumptions before they propagate into structuring.

Phase 2 — Analysis & Modeling

Collateral Analysis

Empirical loss curves from 500+ historical deals

Match the pool against EDGAR shelf performance data. Build empirical CDR, severity, and CNL curves from comparable deals, with vintage-over-vintage overlay and stress multipliers.

EDGAR Deal Matching

Access a database of 500+ historical deals and 100M+ loan observations from EDGAR filings. Graam matches the current pool against comparable shelves by issuer, vintage, and collateral type.

Vintage Comparison Across Issuers

Overlay performance curves across vintages and issuers to identify relative value and emerging trends. Compare CDR, severity, and CNL trajectories side by side.

CDR / Severity / CNL Milestones

Extract cumulative net loss, default rate, and severity milestones at 12, 24, 36, 48, and 60 months of seasoning. Use these benchmarks to calibrate forward projections.

Bull / Base / Bear Scenario Comparison

Apply stress multipliers to base empirical curves to produce bull, base, and bear scenarios. Compare projected performance across assumptions to size credit enhancement.

Historical Performance

Time series from EDGAR and agency datasets

Access period-by-period performance data for auto ABS deals from EDGAR and residential mortgage data from Freddie Mac SFLLD. Analyze CDR, CPR, severity, and delinquency trends across vintages and issuers.

EDGAR Time Series

Period-by-period CDR, VPR, DQ60, DQ90, and CNL for auto ABS shelves. Filter by issuer, vintage, and deal to build custom benchmarks and track performance over time.

Freddie Mac SFLLD Vintage Cohorts

Loan-level performance data from Freddie Mac's Single Family Loan-Level Dataset. Analyze default and prepayment behavior by origination vintage, FICO band, LTV bucket, and geography.

S-Curve Analysis

Prepayment rates by refinancing incentive bucket. Visualize the classic S-curve relationship between rate incentive and voluntary prepayment speed across vintages and loan types.

Sponsor / Shelf-Level Queries

Query performance at the sponsor or shelf level. Compare how different issuers' deals have performed across market cycles and identify platform-specific risk factors.

Model Training

Build default models on historical data

Train GAM and logistic regression models on loan-level historical performance data. Generate feature importance rankings, out-of-sample validation metrics, and model-derived CDR curves for projection.

Auto-Derived Cohorts from Tape Characteristics

Graam segments the tape into cohorts based on FICO, LTV, term, and geography. Each cohort is matched against historical performance data to calibrate default probabilities.

GAM and Logit Model Types

Train generalized additive models (GAMs) for nonlinear feature relationships or logistic regression for interpretable coefficient-based models. Both produce loan-level default probabilities.

AUC Validation (Minimum 0.65 Threshold)

Every model is validated on an out-of-sample holdout. Models below the 0.65 AUC threshold are flagged, and Graam surfaces the validation metrics so you can assess discriminatory power.

Model-Derived CDR Vectors for Cashflow Projection

Convert loan-level default probabilities into aggregate CDR vectors that feed directly into the cashflow engine. Bridge the gap between statistical models and waterfall execution.

Rating Analysis

S&P criteria, stress testing, and implied ratings

Classify the pool against S&P methodology (prime, nonprime, subprime). Apply rating-level stress multiples to base CNL. Size credit enhancement to target ratings with minimum CE floors.

S&P Pool Classification

Classifies the pool as prime, nonprime, or subprime based on collateral characteristics. Each classification carries different base loss assumptions and stress multiples — AAA stress ranges from 3.5× for standard pools to 2.0–3.5× for subprime, with minimum CE floors at each rating level.

Stress Multiples & CE Floors

Apply rating-level stress multiples to base CNL: AAA 3.5–5.0×, AA 2.5–3.5×, A 2.0–2.5×, BBB 1.5–2.0×. Minimum credit enhancement floors of 4.0% (AAA), 3.2% (AA), 2.4% (A), and 1.6% (BBB) apply regardless of modeled losses.

Implied Ratings & Pass/Fail Tables

Once the waterfall runs, Graam calculates implied ratings for each tranche by comparing credit enhancement against stressed CNL at each rating level. Produces a pass/fail table showing whether each tranche clears AAA, AA, A, and BBB thresholds.

Phase 3 — Structuring & Output

Deal Structuring

Extract from prospectus or build from specification

Parse a prospectus to extract the capital stack, tranches, waterfall rules, and triggers into a validated deal model. Or define a structure from scratch and iterate in natural language.

Prospectus Extraction

Parses the prospectus to extract the capital stack, tranche definitions, coupon types, subordination levels, and structural triggers into a validated deal model. Supports sequential, pro-rata, and hybrid structures.

Iterative Validation

Every structure is submitted to the cashflow engine for validation. The agent checks WAL ranges, payment priority, trigger fire-points, and cashflow conservation. If validation fails, it diagnoses the issue, modifies the deal model, and re-runs until the structure passes — or surfaces the discrepancy for review.

Deal Model Modification

Modify existing structures in natural language — adjust coupon rates, resize tranches, add mezzanine classes, or change waterfall rules. The agent re-assembles the deal, recomputes OC and certificate balances, and re-validates against the engine.

WAL Validation

Validates weighted average life for each tranche across scenarios. Ensures WAL targets are met under base, stress, and severe assumptions, flagging tranches that extend or contract beyond acceptable ranges.

Cashflow Projection

Deterministic waterfall execution across scenarios

Run CDR/CPR/severity scenarios through the open-source Flows engine. Produce tranche-level yield, WAL, duration, and discount margin. Solve for breakeven CDR at target yields.

Waterfall Engine

Executes payment waterfalls exactly as specified — sequential, pro-rata, and hybrid structures. Supports interest and principal priority rules, OC and reserve triggers, writedown allocation, excess spread mechanics, and clean-up calls.

Bond Analytics

Computes yield (IRR), weighted average life, modified duration, and discount margin for each tranche at any given price. Supports price-to-yield and yield-to-price solving, breakeven CDR analysis, and multiple day count conventions.

Collateral Projection

Projects defaults (CDR), prepayments (CPR/ABS), and severity from scalar values, ramp vectors, or month-by-month arrays. Supports multi-scenario analysis with configurable stress multipliers applied to base assumptions.

Deal Specification

Deals are expressed as structured JSON — tranches, coupons, waterfall rules, triggers, and assumptions in a single definition. The DSL supports the full range of ABS and RMBS features: multi-class stacks, floating-rate coupons with SOFR spreads, and time-varying vectors.

IC Reports

Investment committee packages, end to end

Built on Graam's extraction, modeling, and scenario engine. Upload a prospectus and loan tape — Graam produces a complete IC report covering deal structure, collateral analysis, loss projections, waterfall analytics, and breakeven analysis.

Every IC report includes

Deal Structure Extraction

Parses the prospectus to extract the capital stack, tranche definitions, coupon types, subordination levels, and structural triggers into a validated deal model.

Collateral Analysis

Stratifies the loan tape, computes weighted averages, and identifies concentration risks that drive loss assumptions and structuring decisions.

Loss & Prepay Forecast

Builds CDR, CPR, and severity curves from empirical data. Projects collateral performance under base, stress, and severe scenarios.

Breakeven & Credit Enhancement

Breakeven CNL per tranche, cushion multiples against projected losses, and breakeven CDR solving at a target yield when a purchase price is specified.

Waterfall & Tranche Analytics

Full cashflow waterfall under each scenario, producing tranche-level yield, WAL, modified duration, and discount margin.

Historical Benchmarking

Empirical performance from matched cohorts — CDR, severity, VPR, and CNL milestones at 12, 24, 36, 48, and 60 months of seasoning.

RMBS (NQM, Agency, CRT)

Adds GSE Loan Performance Datasets as the empirical data source, with S-curve analysis showing prepayment rates by refinancing incentive bucket. NQM stress multipliers are calibrated to documentation type (bank statement, DSCR, full-doc) and applied across base, stress, and rate shock scenarios.

GSE Performance DataS-Curve AnalysisNQM Stress MultipliersRate Shock Scenarios

Auto Loan ABS

Adds EDGAR shelf performance as the empirical data source, with vintage-over-vintage comparison across major auto shelves. Loss curves are calibrated to the issuer's historical experience and current macro environment, projected under base, stress, and severe scenarios.

EDGAR Shelf DataVintage ComparisonIssuer BenchmarksMulti-Scenario