Hope this can help frenz who prep for Bloomberg NG interivews this year for 2027 NG or 2026 NG.

Story Begins~~

Bloomberg did not replace its old interview style in 2026. The same core topics were still there. What changed was the layer around them. More 2026 reports moved from a normal coding problem into state, scale, failures, or system behavior (I think we can use this signal to assume the problem type for 2027 NG interviews :).

What stayed the same 👀

1. Classic data structures still mattered most (๑ᵔ⤙ᵔ๑)

Both years repeatedly used strings, hash maps, linked lists, trees, graphs, grids, intervals, stacks, queues, and heaps. Problems such as Number of Islands, Word Search, Decode String, multilevel linked-list flattening, and constant-time add/remove/random showed up across the two cycles.

2. The follow-up often mattered more than the first answer

In both years, a correct first solution was only the start. Interviewers asked for better time or space complexity, fewer library shortcuts, different input rules, or one more product requirement.

Candidates also had to explain why the new version still worked.

3. Communication was part of the technical score

The strongest reports followed the same rhythm: clarify the problem, explain the plan, write the code, test it, and state the complexity. Candidates lost time when they coded before agreeing on the rules or could not explain a pointer change, an edge case, or a data-structure choice (based on candidates reports )).

4. Resume and project questions deep dive

Both cycles included deep questions about one project. Interviewers asked why a design was chosen, what failed, how the candidate handled unclear requirements, and what they would change now.

These questions appeared in early technical rounds, HR, and manager interviews.

5. There was no single fixed loop

A five-conversation path appeared often: one technical screen, two more technical rounds, HR, and an engineering-manager interview. But it was never a rule. Some candidates had campus rounds, some had a virtual superday, and some had interviews spread across several weeks.

What stood out more in the 2025 reports

The 2025 reports more often read like classic coding interviews. A candidate received one or two recognizable problems, solved them under time pressure, handled edge cases, and explained complexity.

Common shapes included:

• strings and parsing;
• islands, tree traversal, and grid search;
• linked-list and pointer work;
• parentheses, stacks, and duplicate removal;
• intervals and meeting rooms; and
• rotated arrays, substrings, and other standard medium problems.

Design questions: Candidates reported browser history, Top-N news, large-data questions, and system design in later London rounds. Still, those questions felt more dependent on the location or interviewer.

What stood out more in the 2026 reports

The classic problems did not disappear. Multilevel linked lists, RandomizedSet-style design, trees, graphs, strings, and intervals all returned. (2026 sounds harder tbh..

The clearer change was the amount of state around the problem. Candidates reported tasks such as:

• counting active users in a rolling time window;
• updating stock prices and returning the current top K;
• collecting document enrichments from an event stream;
• publishing partial results after a timeout;
• tracking latest, maximum, and average prices in constant time; and
• building small stateful systems such as autocomplete, to-do lists, card games, and Connect Four.

The follow-ups also sounded more operational. Interviewers added duplicate events, out-of-order timestamps, stale heap entries, timeouts, idempotency, sharding, frequent queries, partial results, and larger traffic.

The final stages also had a wider range. Some manager rounds were still only resume and behavioral discussion. Others covered Kafka, reconciliation, legacy-service changes, reliability, or debugging a failing service.