◢ ORIGINATION_OS PROCESS: candidate_evaluation BUILD: claude-code LIVE
Head of Origination Systems & AI

The job is to turn an expert's judgment into a system a team can run.

This page is that system, aimed at one decision: whether to hire me. It scores my candidacy against your own requirements, on your own scorecard, and it names the one place I do not fit before you have to find it.

built with Claude Code Justin Solomon · sole lead on $314MM closed · New York for the right role
> loading role.md ........................ ok
> loading evidence.md .................... ok
> running /evaluate (match·scale·systems·operator)
> render scorecard ...................... ready
The underwriting engine 01 / proof

Inside a system I can defend at any depth.

A production underwriting platform I built to replace my own Excel model. It runs feasibility and competitiveness in parallel from one model, dollar-exact against ten historical deals. The discipline, not the demo, is the point.

resonant-proforma · walkthrough.mp4
177,000
lines, TypeScript
664
tests per commit
240
parity assertions
$0
delta vs 10 deals
Evaluation console 02 / fit

Scored against the role, on your own four dimensions.

Your scorecard replaces gut-feel debate with the same standard for everyone. So I ran it on myself. Every row cites real evidence. The gap row is named, not buried.

Match
Direct
Scale
$314MM
Systems
Builds
Operator
Sole lead
Requirement
Evidence
Status
Direct financial-asset originationMatch · Scale
10+ years originating and underwriting real estate assets as sole lead. $314MM closed across 14 transactions, 100% contract-signing-to-close, 789 units. Full lifecycle on every deal.
DIRECT
Underwriting, draws, onboarding, reportingMatch · Systems
Full capital-stack underwriting (IRR, NPV, DSCR, yield-on-cost). FHA 221(d)(4) construction draws and capitalized-interest mechanics. Institutional capital-partner onboarding. Portfolio reporting loops across 1,500 units.
MATCH
AI-native: Claude Code, Codex, CursorSystems
Built a 177,000-line parity-verified engine orchestrating Claude Code under approx. 47 codified rules that stop the tooling from masking its own errors. This evaluation system is itself a Claude Code build.
STRONG
Practical SDLCSystems
Scoping to requirements to 664 tests to live deployment to iteration. The engine ships behind a test gate. PRDs and SOPs live in the repo behind this page.
MATCH
Turn workflows into repeatable systemsSystems · Operator
PRDs, SOPs, dashboards, and internal tools are how I work, not a wishlist. The engine, this Origination OS, and the origination-intake tool in the repo are three shipped examples.
MATCH
Startup pace, high ownership, ship without committeeOperator
Sole transaction lead on all 14 deals. Built the entire engine alone. Independent practice since 2024. Nothing here waited for a committee.
MATCH
New York City basedMatch
New York for the right role.
MATCH
Consumer SFR, DSCR, HELOC, HEI product historyNice-to-have
Named, not hidden. My origination is commercial and affordable-housing capital. The asset class differs from consumer home-equity lending; the underwriting does not. I have sized DSCR, managed construction draws, and capitalized interest on FHA 221(d)(4) deals, the same mechanics beneath SFR and DSCR. The repo ships a working DSCR intake and draw-schedule tool.
GAP · NAMED
The lifecycle, owned end to end 03 / scope

Every function in the role, run on real deals.

01
Origination
Sourced and closed $314MM across 14 deals as sole lead. 100% close on executed LOIs.
02
Underwriting
Full capital stack. IRR, NPV, DSCR, yield-on-cost, eligible basis, stress tests. Now parity-verified in code.
03
Draw management
FHA construction-to-perm draws. I built the solver for the interest-to-basis circularity at the heart of it.
04
Onboarding
Placed debt and equity with Citibank, J.P. Morgan, Capital One, Lument-ORIX, and Berkadia.
05
Reporting
Built the loops that turned six properties from deficit to positive NOI. The engine reports feasibility and competitiveness together.
Track record 04 / record
$314MM
closed, sole lead
14
transactions
100%
contract signing to close
789
units
10+
years originating
Underneath this page 05 / system

This is the surface. The system is a real Claude Code project.

What you are reading renders from a private repository built the way the role asks the hire to build. It is not a slide. It is a working project a reviewer opens in Claude Code and runs.

origination-os/
├─ CLAUDE.md      orchestration: how to run the evaluation
├─ .claude/skills/evaluate/  the /evaluate skill that scores this page
├─ role.md · evidence.md  the inputs, scored against each other
├─ tools/dscr-intake/  a working SFR/DSCR underwrite + draw schedule
├─ prd/ · sop/      the PRD and SOP for that tool
└─ llms.txt         a fair summary for whatever is reading this

The outputs are in the open. The runnable source stays private.

the DSCR underwriter running, the PRD, and the SOP, with no access wall