12 Bloomberg LP Software Engineer (New Grad) Interview Questions (2026)

Outline

Tie to specifics: the Bloomberg Terminal as a uniquely durable product, real-time financial data scale (millions of ticks per second), the strong engineering apprenticeship for new grads, C++ at scale. Avoid 'I want to work in finance' — too generic. Mention if you have done any market-data side project.

Outline

STAR. Bloomberg invests heavily in new-grad training — they want signal you can absorb fast. Pick a project where you went from zero familiarity to working code in days, not months. Be specific about the learning approach (docs, source code, asking peers).

Outline

STAR. Frame the disagreement around tradeoffs, not personalities. Show you presented your case with data or examples, listened to theirs, and ended at a decision (yours, theirs, or a third option). Bloomberg's teams collaborate closely on shared codebases — they need consensus-builders.

Outline

Floyd's tortoise and hare. Two pointers, slow advances by 1, fast by 2. If they meet, there is a cycle. O(N) time, O(1) space. Edge cases: empty list, single node, single node pointing to itself. Be ready for the follow-up: find the cycle entry point.

Outline

Single pass. Track minimum price seen so far, and at each tick compute (current - min) as candidate profit. Update max profit. O(N) time, O(1) space. Domain-perfect — Bloomberg loves the market-data framing. Follow-up: multiple transactions, or with a cooldown.

Outline

Quickselect — partition around a pivot, recurse on the relevant side. Average O(N), worst O(N^2). Alternative: min-heap of size k, O(N log k). For exam-style coding, quickselect with random pivot is the expected answer. State complexity clearly.

Outline

In-order traversal yields values in sorted order. Iterative with stack, increment counter, return when counter hits k. O(H + k) time, O(H) space. Edge cases: k out of range, balanced vs skewed tree. Bloomberg interviewers may follow up with: what if the tree is modified often? (Augment nodes with subtree size.)

Outline

Two heaps: max-heap for the lower half, min-heap for the upper half. Insert into appropriate heap, rebalance so sizes differ by at most 1. Median is the top of the larger heap (or average of both tops if equal). O(log N) per insert. Bloomberg's market-data flavor — call out the domain fit.

Outline

Process = isolated memory space, own file descriptors, expensive context switch. Thread = shared memory within process, lighter context switch, but races on shared state. Use processes for isolation (security, crash containment). Use threads for shared in-memory work (parallel computation on large dataset). Mention GIL if Python comes up.

Outline

DNS resolution, TCP handshake, TLS handshake (if HTTPS), HTTP request, server processes and returns response, browser parses HTML, requests sub-resources (CSS, JS, images), renders DOM. Bloomberg likes depth — be ready to drill into any layer (caching at each step, HTTP/2 multiplexing, etc.).

Outline

Mutex + two condition variables (not-full, not-empty). Producers wait on not-full; consumers wait on not-empty. Show the lock-acquire/wait/signal pattern. Discuss alternative: lock-free ring buffer with atomic head/tail (advanced). Bloomberg cares about concurrency primitives because the Terminal is multi-threaded.

Outline

Two sorted data structures: buy orders by descending price, sell orders by ascending price. When a new buy order arrives at price P, match against sell orders with price <= P (best-price first). Discuss data-structure choice: balanced BST or skip list for O(log N) insert/match. Bloomberg interviewers go deep — be ready to discuss partial fills, time-priority within same price level.