Domain — syndicated loans / books-and-records
Company- and market-specific notes. Keep System design generic. Use this folder when the loop is a platform that captures agent books and records and builds workflows on top.
Public framing (not an internal wiki): a New York company the large agent banks built so the syndicated loan market can stop running on email, fax, and retyping. It pulls loan data from the agent’s own books over APIs, normalizes it, and only shows each lender what they’re entitled to see.
Core product: Digital Data Platform. Workflows on top: reconciliation, cashless rolls, letters of credit, loan voting. Coverage on the order of 11,000+ facilities and about $4.8 trillion in active commitments. Pitch words: transparency, efficiency, velocity. Wrong data is a trust failure, not a UI bug.
60-second recap
Golden source from agent books, then workflows: VRM (reconciliation) and loan voting. Scale: 11,000+ facilities, ~$4.8T. Wrong data is a trust failure, not a UI bug.
What problem they exist to solve
The broadly syndicated loan and private credit markets grew huge while ops stayed analog:
- Notices and positions arrive by email/fax; people rekey them
- Labels and structures don’t match across agents
- Lenders can’t easily tie out “what I think I own” vs “what the agent’s books say”
- Amendments and votes are spreadsheets, calls, and faxes under a deadline
Bet: if you capture the agent golden source once, digitally, you can build every workflow (reconcile, vote, cashless roll, LC exposure) on the same permissioned data instead of scraping the market.
Credibility is the product.
Products you’d actually ship against
Digital Data Platform (DDP)
- Direct connectivity to agent servicing systems via API (not email scrape as the architecture)
- Normalize facility / position / cash-flow / notice data
- Consume via UI or API
- Strict permissioning: a lender only sees what they’re entitled to
Mock: Agent ingest.
Reconciliation (VRM)
Electronically match lender positions to agent golden-source at a point in time. Show the digital notice that caused the break. Workflow + correctness, not a pretty dashboard.
Mock: Reconciliation.
Loan voting
Digitize amendment / consent voting. Positions come from agents’ systems (settled, by legal entity), not last UI cache. Agents set terms and deadlines; lenders vote, including multi-fund.
Mock: Loan voting.
Other workflows (know they exist)
Cashless roll (keep history when a loan is refinanced). Letters of credit (contingent exposure plus funded loans). Don’t design these unless asked.
Scale (use latest, don’t mix years)
Prefer 11k facilities / $4.8T active commitments. Users are bank operations people, not millions of consumers. Pain is burst at deadline, hot facility, multi-tenant leak.
SDS problems that transfer
| SDS mock | Domain analog |
|---|---|
| Webhook | Agent/provider push. Ack vs apply. |
| Rate limiter | Noisy-neighbor tenants; don’t hammer the agent hose. |
| Payment | Provider cannot join your txn; in-doubt. |
| Robinhood | Live progress vs settled position / official vote. |
Also: vocabulary, NFR defaults, proof stories.