A multi-part live-coding problem: resolve each company to its topmost parent in a parent/child hierarchy, aggregate loans onto that top-level company, then match incoming transactions to the right loan. The challenge is reading comprehension and clean map manipulation over long boilerplate, not algorithms.

Requirements - You are given a mapping of parent company → list of child companies, e.g. AA: [BB, CC], DD: [AA]. A company can itself be the child of another parent, so hierarchies chain (CC → AA → DD). - Part 1 — Loan aggregation: for each customer's loan, resolve the company on the loan to its topmost parent (follow the parent chain until a company has no parent) and record/aggregate the loan under that top-level company. Print each loan's resulting info. - Part 2 — Transaction matching: given a stream of transaction requests (each carrying a company / identifier), find which Part-1 loan each transaction matches. - Follow-up: how would you productionize this, or handle it as a continuous stream of requests. - The parent-company mapping may be nested through many levels, but the hierarchy is acyclic. Edge cases - A parent company may have no child companies — guard map lookups before updating. This is the single most common bug. - Be careful updating the nested map; missing keys are where candidates stall. Notes - Despite suggestions of topological sort or union-find, no real algorithm is required — walking up the parent chain with a plain map suffices. Candidates who write a generic solution have been told to stop and just use the conditions given. - The prompt and skeleton code are long; budget several minutes just to read. You must use the provided classes and helper functions rather than writing your own. - Appears both as a phone screen (店面 / 电面) and as an onsite coding round; in multi-part variants Part 1 must pass before Part 2 unlocks. - Interviewers often stay silent and give few hints even when you stall. - Confirm the hierarchy contract before coding; the current variant permits arbitrarily deep nesting but excludes cycles. Preparation - Practice walking a parent-pointer chain to its root with a while loop, handling roots (no parent) and guarding against cycles. - Drill building and safely updating nested Map<String, Map<... structures (check-or-create keys). - Rehearse reading a long statement and restating the exact requirements before coding—use only the conditions given, no generic over-engineering.

Loan Aggregation & Transaction Matching (Parent-Company Resolution) — Affirm Interview Question

Requirements

You are given a mapping of parent company → list of child companies, e.g. AA: [BB, CC], DD: [AA]. A company can itself be the child of another parent, so hierarchies chain (CC → AA → DD).
Part 1 — Loan aggregation: for each customer's loan, resolve the company on the loan to its topmost parent (follow the parent chain until a company has no parent) and record/aggregate the loan under that top-level company. Print each loan's resulting info.
Part 2 — Transaction matching: given a stream of transaction requests (each carrying a company / identifier), find which Part-1 loan each transaction matches.
Follow-up: how would you productionize this, or handle it as a continuous stream of requests.
The parent-company mapping may be nested through many levels, but the hierarchy is acyclic.

Edge cases

A parent company may have no child companies — guard map lookups before updating. This is the single most common bug.
Be careful updating the nested map; missing keys are where candidates stall.

Notes

Despite suggestions of topological sort or union-find, no real algorithm is required — walking up the parent chain with a plain map suffices. Candidates who write a generic solution have been told to stop and just use the conditions given.
The prompt and skeleton code are long; budget several minutes just to read. You must use the provided classes and helper functions rather than writing your own.
Appears both as a phone screen (店面 / 电面) and as an onsite coding round; in multi-part variants Part 1 must pass before Part 2 unlocks.
Interviewers often stay silent and give few hints even when you stall.
Confirm the hierarchy contract before coding; the current variant permits arbitrarily deep nesting but excludes cycles.

Preparation

Practice walking a parent-pointer chain to its root with a while loop, handling roots (no parent) and guarding against cycles.
Drill building and safely updating nested Map<String, Map<...>> structures (check-or-create keys).
Rehearse reading a long statement and restating the exact requirements before coding—use only the conditions given, no generic over-engineering.

Requirements

Edge cases

Notes

Preparation

Comments