For CDCs · Farm Credit · USDA · HUD · EB-5 · C-PACE · CDFI · Private Credit

The configurable underwriting engine built inside an active credit operator.

One borrower file from application through closing, servicing, and review — and a new loan program deployed as a configuration, not a build.

Source-cited AIReview- and examiner-readyLive in 60 days

$1.5B+
Active program-credit deals
50+
Deals currently underwriting
60 days
Typical live window
FIGURE · CORE PROGRAMS · CITE-OR-DECLINEL-2418 · MERCER MFG.

§ Product proof · Source to audit

From source evidence togoverned credit judgment.

Real, redacted product surfaces. The source document, extracted fact, analyst decision, and audit event stay in the same borrower file.

01Source document
Redacted CORE document viewer showing a borrower financial statement with extracted source facts in the side panel.

The file stays visible while CORE extracts facts.

02Credit signal
Redacted CORE extraction notes pane showing anomaly notes and analyst review cues for a borrower file.

The model flags what a credit officer would investigate.

03Audit event
Redacted CORE audit log showing a classifier event with evidence and raw payload details.

Every decision lands in the event log with evidence attached.

MORE CAPTURED SURFACESINTAKE INBOXSPREADING + REVIEW DRAWERDEAL-FILE SEARCHCOVERAGE MAPView the platform

§ Programs · Supported

Built for the lenders other vendors call “edge cases.”

01
CDC
SBA 504
Owner-occupied RE
02
Farm Credit
FCS Title I & III
Cooperative ag credit
03
USDA
B&I · CF · REAP · Rural
Rural mission
04
HUD
232 · 223(f) · 221(d)(4)
Multifamily insured
05
EB-5
Regional Centers
TEA · I-526
06
C-PACE
Energy · Water · Resil.
Capital programs
07
CDFI
Cert. CDFI lenders
Mission lending
08
Private Credit
Bridge · ABL · Direct
Bespoke programs

§ The Engine · Program Studio

One underwriting engine.
Every program is a configuration.

Plenty of software moves the file. CORE underwrites it — then carries the decision through closing, servicing, committee review, diligence, and examination without rebuilding the platform around each program.

Your document list.

Collection packages — base plus add-ons — not a hardcoded checklist.

Your credit box.

Underwriting policies built from your own metric, covenant, and scorecard catalogs — versioned, assigned per program, switchable mid-deal when the lender or capital source changes.

Your memo.

Templates define sections, intents, and mandates. There is no fixed memo.

Your industry research.

COREView reports generate from configurable profiles — the same discipline as the memos.

A new loan program is a configuration, not a build

PIPELINE · ROLLING WINDOWSOURCE · CORE PROJECT ACTIVITY

$1.5B+across50+active commercial-credit deals currently underwriting on CORE

C-PACE energy + resiliencyUSDA B&I + REAPEB-5 regional center capitalHUD bridge-to-perm housingBRIDGE interim capital

§ Who We Serve

The segments we serve in depth. One connected decision trail.

One engine, configured by program — packaging, decision, and replay surfaces tuned to each program’s eligibility, review posture, and capital path.

CDC

CDCs

504 packaging, eligibility, and credit memo — built against real SBA examination findings.

View CDC build
USDA

USDA

B&I, Community Facilities, REAP — rural mission, federal scrutiny.

View USDA build
CDFI

CDFI

Mission lending with cert-level reporting and donor-defensible files.

View CDFI build
PRIV

Private Lenders

Bridge, ABL, and direct lending — bespoke programs as configurations, files ready for LP and capital-partner diligence.

View Private Credit build

§ AI · Built for examination

AI in CORE is judged by its citations, not its prose.

These controls answer the question credit teams get after the memo is written: where did this claim come from?

  • Cite-or-decline. If the model cannot ground a claim in a file, rule, or accepted fact, it declines to write it.
  • Analyst-owned decisions. The model proposes; the analyst accepts, edits, or rejects. The human decision is logged separately.
  • Audit-ready snapshots. Model, inputs, output, and overrides are frozen against the loan file — replayable after the decision.
  • The receipt. The analyst assembles the evidence; the receipt proves what the model saw.
  • Organized like a credit department. 23 personas across four divisions — underwriting, credit policy, program policy, operations — route section work to analyst-and-supervisor pairs per program. Every agent action is timestamped, actor-named, and queryable.
CORE AUDIT LEDGER · GENERATION SNAPSHOTFROZEN · EXAMINABLE
decisionFILE PINNED · Tax Return — 2023
modelclaude-sonnet-4-6 · 2026-03-19
skillcredit-analysis-spreading@v3.2.1
candidatesMercerMfg_2023_1120S.pdf · conf 0.94 · PINNED
MercerMfg_2023_K1.pdf · conf 0.81
analystR. Tabori · approved · 14:22:11 EDT
snapshot3f9c-b21e-44a7-c019
FIGURE · AUDIT REPLAYABLEL-2418 · CDC 504

§ Origin · Built in production

Built inside a working commercial-credit shop. Against real examination findings. While the loans were closing.

Every major surface came from pressure inside the work: analysts assembling files, supervisors checking memos, lenders answering examination questions, borrowers waiting on decisions.

SBA, USDA, and conventional bridge are the configurations Waterside runs on its own deals — the product is used on the builder’s book before it is sold.

§ Request demo

Pick the path that matches your program.

Demos are 30 minutes — real deal artifacts, source citations, and the audit trail behind a credit decision. Bring your hardest credit-review question.

Source-cited AI · Human-controlled decisions · Review-ready replay · Live in 60 days