Bloomberg New Grad AI-Assisted Coding Rounds: 3 Experiences
- Company
- Bloomberg LP
- Reports
- 3 interview reports
- Level
- New Grad
- Stages
- Phone screen
In late 2026, Bloomberg new grad loops had two AI-assisted coding rounds: debug existing code, then, in one loop, fix tickets in a larger repo.
At a glance
- Process: entries 01 and 02 had two AI-assisted coding rounds. Entry 01's London loop also had system design (1 hour), HR (30 minutes) and an engineering manager (1 hour), with three interviews in the middle of the loop held as a superday. Entry 03 described one phone-stage round.
- Asked recently: debugging existing code (2 of 3: a ticket booking program of about 100 lines of Python in entry 03, a stock exchange class in entry 02); four tickets in a game matchmaking repo (1 of 3). Entry 01 didn't share its tasks.
- Difficulty: entry 02 rated the interviews hard and got through 2 of the 4 tickets; entry 03 rated its round average and passed; entry 01 did not get an offer.
- Practise first: Debug a Stock Exchange Order Book for entry 02's first round, Debug a Ticket Booking System for entry 03's round, and Resolve Tickets in Multi-Tenant Matchmaking for entry 02's ticket round.
Coverage: 3 new grad software engineering candidates: one in London, and two (one bachelor's, one master's) recruited at campus career fairs · AI-assisted coding rounds, plus one London superday · reports posted November 2026, one giving an October–December 2026 interview window
Other levels: Bloomberg SWE intern AI-assisted coding rounds, Bloomberg experienced engineer AI coding interviews and Bloomberg AI interview reports with no level given.
What came up
The two reports that shared their tasks were both handed code someone else wrote and asked to find what was broken, with AI help available.
| Topic | Reports | Last seen |
|---|---|---|
| Finding and fixing bugs in existing code | 2 of 3 | Nov 2026 (report date) |
| Explaining the reasoning, data structures and runtime behind the fix | 1 of 3 | Nov 2026 (report date) |
| Writing your own test cases | 1 of 3 | Nov 2026 (report date) |
| Working through tickets in a larger repo | 1 of 3 | Nov 2026 (report date) |
| Race conditions between concurrent requests | 1 of 3 | Nov 2026 (report date) |
Questions we found
AI Coding
| Question | Reports | Last seen |
|---|---|---|
| Fix the bugs in a stock exchange order book class | 1 | Nov 2026 (report date) |
| Resolve four tickets in a multi-tenant game matchmaking engine | 1 | Nov 2026 (report date) |
| Debug a ticket booking system | 1 | Nov 2026 (report date) |
3 detailed candidate timelines
01: Two AI interviews in a five-interview London loop
Report date: Nov 16, 2026 · Role: SWE New Grad · Location: London · Result: No offer
The loop
- Five interviews in all: two AI-assisted interviews (listed at one hour; the report doesn't say whether that is each or both together), system design (1 hour), HR (30 minutes) and an engineering manager (1 hour).
- The three in the middle of the loop were held together as a superday, and each invitation came after the interview before it.
- The candidate didn't describe the tasks.
02: An order book to debug, then four matchmaking tickets
Report date: Nov 13, 2026 · Role: Software Engineer, new grad (master's) · Recruited: Campus career fair · Rating: Hard, neutral experience · Result: Not selected
AI-assisted coding round 1: debug a stock exchange class
- The code was a
StockExchangeclass. Its order book was a dictionary from each ticker symbol to a list ofOrderobjects. Selling added an order for a ticker at a given price and quantity. Buying took a quantity and filled it from the cheapest orders. - The job was to find and fix the bugs. The candidate had to come up with their own test cases, explain the reasoning behind each change, and talk through the data structures and the running time.
- The bugs they found:
- nothing checked whether the ticker was in the book before using it;
- when a buy took only part of what was on offer, orders were not removed from the book correctly;
- the method that generates order IDs never moved on to the next number;
- buy and sell accepted zero, negative and other invalid input instead of rejecting it;
- prices were stored as floats, which can cause rounding errors with money.
- They also raised a speed problem: the sell method sorted the whole order list on every call. Keeping each ticker's orders in a min-heap (a structure that always hands back the cheapest order first) avoids the repeated sort.
AI-assisted coding round 2: tickets in a matchmaking repo
- This time the code was a full repository: a multi-tenant game matchmaking engine. Multi-tenant here means several separate game arenas share the system. Players are paired with whoever joined earliest at the closest skill level.
- There were four tickets. The candidate only had time for two:
- A player with an unusual skill level never got a match. If everyone else was close in skill, someone much stronger or weaker waited forever. Their fix let the skill band (the allowed skill gap, set in a method called
effective_skill_band) grow the longer a player waited. - One player could land in two games at once. Two requests could both see the same waiting player before either recorded the match. Their fix wrapped the call that finds a waiting player and the insert call in a lock, so only one request at a time can run that pair of steps.
- A player with an unusual skill level never got a match. If everyone else was close in skill, someone much stronger or weaker waited forever. Their fix let the skill band (the allowed skill gap, set in a method called
- Their tip for this round: get the AI to describe the system first (its main objects, the components and how requests travel through them) and only then start editing. They said this helps you understand an unfamiliar codebase better.
Looking back
- The bottleneck in round 2 was reading speed: a lot of code and ticket text had to be understood quickly before any fix, and two tickets were all the time allowed.
- Their overall verdict was lukewarm, and they did not feel at home in this AI-assisted format.
03: Bugs in a ticket booking program, and a pass
Report date: Nov 8, 2026 · Interview window: October–December 2026 · Role: Software Engineer, new grad (bachelor's) · Recruited: Campus career fair · Rating: Average difficulty, positive experience · Result: Passed
Technical phone screen, round 1: AI debugging
- The candidate passed this first round, which was a debugging exercise rather than an algorithm problem.
- The code simulated booking tickets: a Python program of a little over 100 lines that already contained bugs. Using AI to help track them down was allowed.
- What they heard from others, not what they saw: roughly five bugs in total, and someone who found four still passed. That fits their own impression that the bug count was not the whole score.
- In their view, two other things were being judged: how clearly they talked the interviewer through the work, and how they used the AI.
Patterns across the reports
- Two AI-assisted rounds. Entries 01 and 02 both had two AI-assisted coding rounds; entry 01's loop also included system design, HR and a manager, while entry 02 did not get past the AI rounds.
- A debugging round first. In entries 02 and 03 the first AI-assisted round was finding bugs in existing code: a ticket booking program in entry 03 and an order book class in entry 02. In both, explaining the work counted alongside the fixes: entry 02 had to justify data structures and running time, and entry 03 felt communication and AI use were judged.
Common questions
What is Bloomberg's AI-assisted coding round for new grads?
In the two reports with tasks it is a code-reading round, not an algorithm puzzle: you get existing code with bugs (entries 02, 03) or a repository with tickets (entry 02) and fix it with AI help available.
Which AI tool does Bloomberg let you use?
None of the three reports named the tool. Entry 03 said only that AI help was allowed; entry 02 talks about asking the AI to explain the system before editing.
Do you have to fix every bug or ticket in Bloomberg's AI rounds?
Not always. Entry 03 passed and had heard that one candidate passed with four of about five bugs. Entry 02 got through two of four tickets and was not selected, but the report does not say which round decided it.
What do Bloomberg interviewers look at besides the fix?
Entry 02 was asked for its own test cases and for the reasoning, data structures and running time behind each change. Entry 03 believed clear communication and how the AI was used both counted.
What to practise
- Debug a Stock Exchange Order Book. Bugs in a small Python or Java exchange that keeps an order book of sell orders, with an AI assistant available: the same task as entry 02's first round (1 of 3 reports).
- Resolve Tickets in Multi-Tenant Matchmaking. The two tickets entry 02 described (a player who never gets matched, and one player in two games at once) plus two more, in a matchmaking engine with an AI assistant available (1 of 3 reports).
- Debug a Ticket Booking System. Bugs in a small Python booking system, with an AI assistant available: the same task as entry 03's phone screen, and the same kind of task as entry 02's first round (2 of 3 reports; the third didn't describe its tasks).
Keep practicing
- Debug a Stock Exchange Order Book: the order book project behind entry 02's first round.
- Resolve Tickets in Multi-Tenant Matchmaking: the matchmaking ticket project behind entry 02's second round.
- Debug a Ticket Booking System: the Bloomberg project closest to entry 03's round.
- Bloomberg project coding bank: every Bloomberg codebase exercise on FastPrep (sign-in needed).
- Bloomberg interview guide: the company page with the full question list.
- Bloomberg hiring insights: recent interview activity at Bloomberg.
- Experience feed: more firsthand interview reports.
AI coding rounds at other companies
- Meta: Meta AI coding interview: two onsite rounds where you finish a hard algorithm inside a mostly written codebase with an AI model, then face stress-test suites.
- Google: Google AI coding interview: an online assessment with HackerRank's built-in assistant, and a live round on a multi-file project where using AI was optional.
- Stripe: Stripe AI coding interview: experienced engineers kept getting the same transaction rule parser, built step by step with a built-in model.
- Stripe interns: Stripe intern AI coding round: ten intern accounts of the same AI-assisted round.
- Rippling: Rippling AI coding interview: HackerRank's built-in AI on Rippling's familiar problems, plus an HR codebase at the onsite.
- DoorDash: DoorDash AI coding interview: AI CodeCraft: build and run several small services live with your own AI tool.
- LinkedIn: LinkedIn AI coding interview: one long, multi-part task in an IDE whose agent edits files directly.
- Shopify: Shopify AI coding interview: pair programming with your own AI on Shopify's well-known question set.
- Datadog: Datadog AI coding interview: bring any AI tool and share your screen; the tasks are easy, and the follow-ups decide the round.
- Capital One: Capital One AI coding interview: new grad Power Day agentic round: fix a bug, then add a feature to a large full-stack app with Claude Code.
- xAI: xAI AI coding interview: new grad take-homes and onsites where you build with AI.
- Adobe: Adobe AI coding interview: AI showed up as one step in a director round and in a take-home; the rest stayed classic.
- Snowflake: Snowflake AI coding interview: design and build a rate limiter or a retrying task queue while the AI writes the code; one AI Engineer onsite was in person with your own laptop.
- Snap: Snap AI coding interview: a backend-only chat service from an empty Python project with Claude or Codex, or a feature in an existing codebase.
- Ramp: Ramp AI coding interview: build inside a repo Ramp provides; one repo logged every prompt for the interviewer to read.
- Anthropic: Anthropic AI coding interview: review four pull requests, then improve one with Claude Code.
- Apple: Apple AI coding interview: CoderPad with a sidebar agent, alongside Apple's other recent interviews.
More companies with AI rounds on FastPrep:
- Amazon: Amazon interview report: an AI-assisted debugging OA (1 experience).
- Atlassian: Atlassian interview report: project coding and AI tool use (11 experiences, two parts).
- Circle: Circle interview report: three AI rounds: building, debugging and PR review (1 experience).
- Volta: Volta interview report: a deliberately vague data task with a token budget (1 experience).
Across companies: 2026 AI Coding OAs at 15 Companies: 32 Reported Problems and 2025–2026 AI Coding Phone Screens and Onsites at 43 Companies: 76 Reported Problems.
Sources: Anywhere useful interview info can be found.
Comments
0No comments yet — be the first to share what you know.