Graam is your credit analyst. It reads your deal documents, loan tapes and data rooms, joins them with public and agency data, and does the work a desk hands to an analyst: the IC memo, the diligence on a tape, the structure sized to rating criteria, the prepayment report, the trade idea. Every number it writes is traceable to the table it came from.
1. What it reads
Whatever is in your data room. Prospectuses, term sheets, presales, indentures, remittance reports, loan tapes, servicer files, your own research. Graam indexes each document, reads the tape's columns from the headers and the values, and organizes everything by deal, issuer and collateral, so a question about one deal is answered from that deal's own documents and a question across the book is answered across the book.
2. What it already knows
The public and agency data a desk otherwise pays for or builds itself is preloaded and joined to your deals:
- County property records: assessed and market values, owner of record, mailing address, sales.
- Mortgage originations from HMDA, by lender, county and purpose.
- Agency credit-risk-transfer loan performance from Freddie Mac and Fannie Mae, monthly.
- CMBS loan performance, deal by deal.
- SEC filings and economic series from FRED and the EIA.
That is what makes diligence start the moment a tape lands: the loans are matched to the county, the servicers to the agency book, the market to the same data the agencies publish.
3. What it produces
Deliverables, not chat. An IC report with every figure cited to its page. A diligence report on a tape. A capital structure sized to S&P or Fitch criteria with break-evens and yields. A prepayment or performance report for a pool, a deal or a program. A deal's terms as data, read from its own prospectus. Research on an individual loan: its property, its borrower profile, its history, its servicer. Each is a skill: a fixed procedure over typed primitives, so the arithmetic is the same every time and the model's job is to read, judge and write. Your desk's own process can be saved as a skill too, and rerun on every new file.
4. Three things it did this month
Diligence on a loan tape. Three hundred and two loans, one instruction: run due diligence. Graam matched 301 of them to exactly one county parcel, flagged 13 appraisals at more than one and a half times the county's market value, found 10 owner-occupied loans whose owner of record gets mail elsewhere and 4 whose owner is a company, scored the servicers against the agency book, sized a structure to S&P criteria, and named the loans to look at with a reason for each. Two minutes. The tape was synthetic with the problems planted in it, and it surfaced every property plant and dismissed the one planted trap.
A CLO tied to the manager's cashflows. From the indenture and the collateral file, Graam built the deal and ran it against the manager's own reference cashflows under three scenarios. Rated interest tied to within one percent on every class and to a tenth of a percent on the senior, with the remaining differences attributed line by line to stated conventions in the indenture rather than to the model.
A CRT performance report. For a credit-risk-transfer deal, Graam reads the agency's monthly loan-level file, reports delinquency, prepayment and losses by activity month, and reconciles the reported loss to its components to the cent. The same report runs for any STACR or CAS deal, and for the program as a whole.
5. Where it runs
In our cloud, or in yours. Your documents and data are scoped to your firm and are never shared with other customers. The skills you write are yours to read, change and run.
6. What comes next
One post at a time, each a single question Graam answered and how, ending where this one ends. If you work in RMBS, CMBS, CLOs or consumer ABS and want the analyst on your desk: sign up, upload a deal you already hold, and ask it something you would ask an analyst.