Agent data ingest
You are interviewing for Engineering Manager. This is a 45–60 minute whiteboard / word-problem. Talk through observations as you go. Draw.
Read this document only. Do not open the interviewer page.
Context
You work at a company whose mission is to make the syndicated loan market less analog. For decades, notices and positions moved by email and fax; people retyped them; labels did not match across agent banks.
The company captures data from agent banks’ own servicing systems and centralizes it so lenders, agents, and service providers share one permissioned view of deal information. Credibility is the product: participants need to believe they are looking at the loan market’s most reliable source of deal data.
Coverage is on the order of thousands of facilities and trillions of dollars in commitments. Users are bank operations people and systems integrating by API — not a consumer social network.
You are not scraping inboxes as the design. Connectivity to agents is meant to be digital.
Problem
Design the system that brings agent-bank loan data onto the platform and makes it usable.
The platform needs facility, position, cash-flow, and notice data. Different agents will not use the same labels or structures. Consumers of the data include an internal UI and external APIs. Each participant must only see what they are allowed to see. The business talks about “real time.”
Banks will be onboarded over time; a new agent going live is a company event, not a silent config change.
What to produce
- Clarify requirements. Ask questions. Write down what you are optimizing for.
- Sketch APIs or main operations (ingest and query).
- Draw a high-level design and walk through an event from an agent to a user read.
- Go deep on the hardest parts.
- Close as an engineering manager: what ships first, who owns what, what you measure, how you talk about risk.
You may assume a cloud environment similar to Azure. You do not need to pick a UI framework.