Company interview guide

Bloomberg Machine Learning Interview Guide (2026)

Classic coding rounds, solid ML and NLP fundamentals and an open-ended model design, usually with two interviewers per round.

Bloomberg's machine learning interview loop has 7 reported stages, from recruiter or hr call to hiring manager, team and hr rounds; typical length: 2-12 weeks; most reports take one to two months.

Roles covered: ML Engineer, AI Engineer, Research Scientist and Data Scientist. The question bank holds 6 questions reported from Bloomberg interviews.

Try a free question from this loop's topics

Interview process length
2-12 weeks; most reports take one to two months
Stages
7
Reports linked to bank questions
7
Bank report dates
Aug 2017–Sep 2024
Questions in bank
6 (0 free)
Last reviewed
October 2026

How the Bloomberg interview loop works

A 15-40 minute recruiter or HR call, then one or two technical phone screens of 45-60 minutes; ML candidates often get one screen on coding and one on ML fundamentals, or a single screen split between a coding problem and a talk about past work. The final stage is two hours to a full day of rounds, each about an hour and often run by two engineers from the hiring team: more coding, an ML or NLP depth round, an ML system or model design round, then conversations with a manager and HR. Research and PhD candidates sometimes present past work. Data science loops in London and Milan in 2025 used a coding round, a Jupyter or pandas data task and a team-lead chat; some data and analyst tracks start with online tests or recorded video answers. Recruiters may send a round-by-round guide, but one 2026 candidate got a different round than the guide listed, so prepare for coding and design either way.

The Bloomberg interview process, stage by stage

01Recruiter or HR call

About 15-40 min by phone with a recruiter or HR; an engineer sometimes runs the first call

A short check on your background, interest in the role, salary expectations and logistics. For ML roles the recruiter usually explains the round types that follow; one 2026 candidate also got a written interview guide listing each round.

What it tests

  • Why Bloomberg and why this team
  • How your experience maps to the role
  • Logistics: location, level, timeline and salary expectations

How to prepare

  • Ask for the type and length of every later round, and whether coding is algorithmic or a practical data-structure task.
  • Have a two-minute summary of your most relevant ML or data project ready.
  • Treat the round guide as a hint, not a promise: be ready for coding and design in any technical slot.

Recruiter or HR intro call in 15 ML, research and data science reports plus 1 AI platform engineer report; five are from 2024-2026. The company's hiring page lists an initial call with HR or an engineer.

02Online tests and recorded video (some data tracks)

Data science and data analysis tracks only: online aptitude or data-skills tests, pre-recorded video answers, sometimes an assessment day

Some data science and analyst openings, mostly in London and Hong Kong, start with online tests (statistics, SQL, Python) and recorded video questions, and one ran an assessment day with a short workflow exercise. A 2024 senior data scientist report had an online test with SQL and an ML case study. ML engineering and AI group reports do not mention this stage.

What it tests

  • Statistics, SQL and Python basics under time pressure
  • Structured answers on markets, news and data trends
  • Suggesting improvements to a process or workflow

How to prepare

  • Ask whether the role sits with an engineering team or a data analysis group; the second is where these tests appear.
  • Practise short, structured spoken answers about financial news before recording video responses.

Online tests, recorded video or an assessment centre in 4 data science reports and 1 research report, only two since 2024; none from ML engineering roles.

03Coding rounds (phone screen and onsite)

45-60 min in a shared editor or HackerRank at the screen, plus one or two onsite rounds of about an hour, often with two interviewers; any language

The most consistent part of the loop for every track. Problems are usually medium-level data-structure questions, often about streams: top-k items over a time window, sliding windows over messages, sorting streaming data. Some interviewers present a practical task with the business context stripped out and then change a constraint so you must extend your code. Data science screens tend to be easier and sometimes include a basic question such as stack versus queue.

What it tests

  • Hash maps, heaps and binary search over streaming or timestamped data
  • Trees and graphs: lowest common ancestor, path enumeration
  • Extending a working solution when the interviewer adds a constraint
  • Clear explanation of complexity and trade-offs while coding

How to prepare

  • Drill the classic stream problems: top-k frequent over a window, hit counters and k largest elements.
  • Get a correct simple version working first; later parts build on it.
  • State assumptions and outline the approach before typing, as the company's guide advises.
  • Prepare even if told there is no algorithmic coding; a 2026 platform candidate still needed binary search.

Coding in 32 of 49 ML, research and data science reports across all years, eight since 2024; streams and top-k recur. The company's hiring page describes a 45-60 minute coding screen and hour-long in-house rounds.

04ML and NLP fundamentals, plus project deep dive

Part or all of a 45-60 min screen, and usually one onsite round; research and PhD candidates sometimes give a talk on past work

Interviewers check that you understand standard models and can talk through your own work. Questions are mostly conceptual: linear versus logistic regression, SVMs, bagging and boosting, overfitting, missing data, and how transformers and attention work. Many rounds start from your resume and probe a past project, a thesis or a paper. Quant-facing teams have added statistical tests and probability.

What it tests

  • Classical ML: regression assumptions, SVMs, random forests, boosting
  • Deep learning and NLP: transformers, attention, vanishing gradients
  • Evaluation: offline versus online metrics, handling overfitting and missing data
  • Explaining the losses, models and choices in your own projects

How to prepare

  • Be able to explain SVMs, ensembles and regression clearly in a few minutes each; they come up again and again.
  • Prepare a crisp transformer walkthrough with shapes and costs, since NLP teams ask for it.
  • Pick one or two projects and rehearse the follow-ups: loss choice, metrics, what failed.

ML fundamentals or a project deep dive in 29 ML, research and data science reports, six since 2024. The company's student page mentions resume and project discussion, but no official page describes ML rounds.

05ML system and model design

About 60 min, at the screen or onsite; open discussion, sometimes followed by coding a few functions

An open problem where you design a model or ML system end to end and then defend loss, metrics and trade-offs. Prompts are often NLP: a spam filter, intent classification, segmenting text with no spaces, grouping chat messages into threads by topic, or a classifier on a mixed categorical and numeric dataset. Senior loops in 2025-2026 also listed general system design and MLOps or API design rounds.

What it tests

  • Framing a vague product need as an ML task
  • Choosing model, loss and metrics, and justifying the trade-offs
  • Data preprocessing and features for text or tabular data
  • Serving, scale and monitoring for senior roles

How to prepare

  • Have a ready plan for a text classification system: data, labels, baseline, transformer model, metrics, errors.
  • Expect follow-ups on any model you name; if you propose a transformer, be ready to explain it in detail.
  • For senior roles, prepare a classic system design as well as an ML design.

ML or system design round in 13 ML, research and data science reports, four since 2024; prompts are mostly NLP. The company's hiring page says design skills are assessed but describes no ML design round.

06Data science case: notebook, pandas and SQL

Data science roles: a live Jupyter or pandas task, or a business case in Python; sometimes SQL

Data science loops add a hands-on data round. Reports from 2024-2025 describe a notebook task diagnosing a dataset, a business case in Python and pandas, and a technical interview mixing pandas, SQL and a short algorithm. ML engineering loops do not report this round.

What it tests

  • Exploring and cleaning a dataset in pandas
  • Writing SQL for data manipulation
  • Turning a business question into an analysis
  • Explaining findings clearly to a non-technical audience

How to prepare

  • Practise a timed notebook workflow: load, check quality, summarize, then answer the business question.
  • Keep pandas group-by, merge and window operations fluent without looking them up.

Notebook, pandas or SQL case in 5 data science reports, three from 2024-2025; none from ML engineering roles.

07Hiring manager, team and HR rounds

Usually the end of the onsite: one or two managers, then HR; about 60 min each where stated

Managers dig into your resume and how you work, and HR checks culture fit and motivation. Teamwork, handling difficult problems and conflict with a manager come up repeatedly. Some candidates met a hiring manager earlier, as a phone screen.

What it tests

  • Teamwork, time management and agile ways of working
  • Handling conflict and difficult challenges
  • Motivation for Bloomberg and fit with the team
  • Depth behind the projects on your resume

How to prepare

  • Prepare short stories on teamwork, a conflict with a manager and a hard technical problem.
  • Know what the team builds and how it uses ML, and connect it to your own work.
  • Bring questions about the team, its technical challenges and its culture.

Manager, team-lead or HR rounds in 17 ML, research and data science reports, five since 2024; behavioral questions are consistent. The company's hiring page says only that the final rounds vary by role.

Questions and themes reported most often

  • Deep dive on a past project, thesis or paper (12 reports, ml and nlp fundamentals, plus project deep dive)
  • Top-k or sliding-window problems over a data stream (6 reports, coding rounds (phone screen and onsite))
  • Linear versus logistic regression and regression assumptions (4 reports, ml and nlp fundamentals, plus project deep dive)
  • Teamwork, conflict and handling difficult challenges (4 reports, hiring manager, team and hr rounds)
  • Notebook or pandas data task for data science roles (4 reports, data science case: notebook, pandas and sql)
  • Explain support vector machines (3 reports, ml and nlp fundamentals, plus project deep dive)
  • Explain transformers and attention (3 reports, ml and nlp fundamentals, plus project deep dive)
  • Bagging versus boosting and random forest splits (2 reports, ml and nlp fundamentals, plus project deep dive)

What Bloomberg asks about

  • Coding & data structures35%
  • ML & NLP fundamentals25%
  • ML system & model design15%
  • Projects & behavioral15%
  • Data & stats case10%

Bloomberg interview prep checklist

  • Solved top-k over a sliding time window and a hit counter without help
  • Can explain SVMs, bagging versus boosting and regression assumptions in under three minutes each
  • Can walk through a transformer block with shapes and attention cost
  • Designed a text classification system with data, model, loss and metrics
  • Rehearsed a deep dive on one project, including what did not work
  • Prepared teamwork and conflict stories for the manager and HR rounds
  • Asked the recruiter for the type and length of every round

Tips

  • Treat coding as the gate: nearly every ML and data science report has at least one data-structure round.
  • Practise stream problems with time windows, heaps and hash maps until they feel routine.
  • Revise classical ML explanations (regression, SVMs, ensembles, overfitting) along with transformers; depth matters more than derivations.
  • Prepare one NLP design end to end, such as spam or intent classification, with metrics and error analysis.
  • Expect two interviewers in many rounds and keep both involved as you talk through your reasoning.
  • Ask the recruiter which rounds you will get, but prepare for coding and design in any slot.
  • Data science candidates should rehearse a timed notebook task in pandas and basic SQL.
  • Have a clear answer to why Bloomberg that connects to finance data, news or the team's product.

Practice the Bloomberg loop

Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.

Bloomberg interview FAQ

How many interview rounds does Bloomberg have for ML roles?

Candidates report 7 stages: Recruiter or HR call, Online tests and recorded video (some data tracks), Coding rounds (phone screen and onsite), ML and NLP fundamentals, plus project deep dive, ML system and model design, Data science case: notebook, pandas and SQL and Hiring manager, team and HR rounds. A 15-40 minute recruiter or HR call, then one or two technical phone screens of 45-60 minutes; ML candidates often get one screen on coding and one on ML fundamentals, or a single screen split between a coding problem and a talk about past work. The final stage is two hours to a full day of rounds, each about an hour and often run by two engineers from the hiring team: more coding, an ML or NLP depth round, an ML system or model design round, then conversations with a manager and HR. Research and PhD candidates sometimes present past work. Data science loops in London and Milan in 2025 used a coding round, a Jupyter or pandas data task and a team-lead chat; some data and analyst tracks start with online tests or recorded video answers. Recruiters may send a round-by-round guide, but one 2026 candidate got a different round than the guide listed, so prepare for coding and design either way.

How long does the Bloomberg ML interview process take?

Reported timelines run 2-12 weeks; most reports take one to two months.

What does Bloomberg ask in ML interviews?

By reported question mix: Coding & data structures 35%; ML & NLP fundamentals 25%; ML system & model design 15%; Projects & behavioral 15%; Data & stats case 10%. The most repeated themes: Deep dive on a past project, thesis or paper; Top-k or sliding-window problems over a data stream; Linear versus logistic regression and regression assumptions.

How should I prepare for the Bloomberg ML interview?

Treat coding as the gate: nearly every ML and data science report has at least one data-structure round. Practise stream problems with time windows, heaps and hash maps until they feel routine. Revise classical ML explanations (regression, SVMs, ensembles, overfitting) along with transformers; depth matters more than derivations.

Are there free Bloomberg interview questions?

LeetML has 6 questions reported from Bloomberg interviews. The stage sections above link the free practice problems for the topics each round covers; the reported-question set itself is in the app.

Related interview guides

Official sources