10 Brex Software Engineer (New Grad) Interview Questions (2026)

Outline

Brex is a corporate card and finance platform for startups and businesses. Pick a specific aspect (the corporate-spend product, the embedded software, the API-first approach). If you have used a corporate card or expense system, ground your answer in that experience. Avoid generic 'fintech is interesting.'

Outline

STAR. Brex handles money — a bug that ships can cost customers real dollars. Pick a story where YOU caught something during review, testing, or production observation. Concrete: what was the bug, what would have happened, how did you catch it, what did you do next. End with what you put in place to prevent the class of bug.

Outline

STAR. Brex values end-to-end ownership. Walk through: how you scoped, what tradeoffs you made, how you got reviews, how you handled feedback, what you measured post-ship. Concrete metrics matter — adoption, error rate, performance. Avoid stories where the team did all the work and you were a passenger.

Outline

Vertical scan: compare character by character across all strings; stop at first mismatch or shortest string end. O(n*m) where n is count and m is average length. Alternative: sort, compare only first and last. Edge cases: empty list, single string, all empty strings. Walk through with a small example.

Outline

Recurse with min/max bounds — every node value must be in (min, max). Update bounds when recursing left (max=node.val) or right (min=node.val). Alternative: in-order traversal must be strictly increasing. Discuss why naive 'left smaller, right larger' check on direct children fails. O(n) time, O(h) space.

Outline

Max-heap of size K keyed by squared distance (skip the sqrt). Push each point, pop if size exceeds K. Result is the heap contents. O(n log K) time. Alternative: quickselect for O(n) average. Edge cases: K > n (return all), K = 0 (return empty). Walk through with a small example.

Outline

Walk transactions, convert each using rates[fromCurrency][toCurrency] or via a base currency. Discuss precision (use integer cents, not floats — Brex cares). Edge cases: missing rate, same-currency transactions. O(n) time. Mention idempotency if reprocessing.

Outline

Sweep line. Events: (start, +amount), (end, -amount). Sort by time, walk events, track running held total, emit per change. O(n log n). Discuss how to handle simultaneous events. Variant: return only the maximum held amount across all time.

Outline

Append-only ledger. Each row is an event with type (capture, hold, refund, reversal), amount, parent_event_id for related events, timestamp, idempotency_key. Discuss why append-only (audit, regulatory) vs mutable rows. Balance computation as a sum or a materialized view. Walk through one transaction lifecycle.

Outline

Client sends idempotency_key. Server stores key + response in a fast store (Redis-style) with TTL. On retry with same key: return stored response. Discuss the race condition (two requests with same key at the same time — use a per-key lock or DB unique constraint). Mention the TTL choice (24h is common). Walk through one cycle.