Company interview guide
Oracle Machine Learning Interview Guide (2026)
Recruiter and hiring-manager screens, then a technical screen and a multi-round loop mixing practical coding, ML and LLM fundamentals, system design and behavioral questions.
Oracle's machine learning interview loop has 9 reported stages, from recruiter or hr screen to behavioral and culture rounds; typical length: 2-8 weeks.
Roles covered: Data Scientist, Applied Scientist, ML Engineer and AI Engineer. The question bank holds 2 questions reported from Oracle interviews.
- Interview process length
- 2-8 weeks
- Stages
- 9
- Reports linked to bank questions
- 2
- Bank report dates
- Jan 2024–Jul 2025
- Questions in bank
- 2 (0 free)
- Last reviewed
- October 2026
How the Oracle interview loop works
A recruiter or HR screen comes first in most reports. Applied scientist and ML engineer candidates (often OCI AI teams) then have a technical screen of about an hour that mixes short ML questions, resume discussion and a coding task, followed by a loop of four or five interviews (one report says about four hours) covering coding, ML or AI fundamentals, system design and behavioral questions. Data scientist candidates usually meet the hiring manager early, then face technical rounds on ML, statistics, Python and SQL, and a final stage of four to six interviews that may include a project presentation. Interns and some early-career roles take an online HackerRank test first. A few data science teams have used a take-home data challenge. The outline rests on short candidate reports; only about eight describe an OCI-style applied scientist or ML engineer loop since 2024.
The Oracle interview process, stage by stage
01Recruiter or HR screen
About 30 min phone or video call with a recruiter or HR
A background and logistics call: your experience, why Oracle, salary expectations and work authorization. Some candidates were asked to rate their own Python and SQL skills, and one was pressed hard on visa status. In applied scientist processes the recruiter often lays out the remaining steps, such as a technical screen and then a loop.
What it tests
- A short, ordered summary of your background and the role you want
- Motivation for Oracle and for leaving your current job
- Salary expectations, location and work authorization
- A quick self-rating of skills such as Python and SQL
How to prepare
- Ask which rounds follow, how long each is and whether they include coding, design or a presentation; formats vary by team.
- Confirm the level you are being considered for before the loop; in two near-identical 2024 reports a senior ML candidate was told after the loop that the opening was junior.
- Have a crisp answer on work authorization ready, since it came up early in several reports.
15 candidate reports (12 since 2024) across data scientist, applied scientist, AI engineer and intern roles start with a recruiter, HR or 30-minute pre-screen call.
02Online assessment (interns and early career)
HackerRank test; reports describe multiple-choice ML and LLM questions, aptitude, coding problems and sometimes a small ML task; one intern test ran three hours
Mainly reported for ML interns, ML assistants and early-career AI engineers, where it comes before any human interview. Content mixes coding problems with questions on ML, LLMs and aptitude. Experienced applied scientist and data scientist hires in our reports usually skip it and go straight to a recruiter call.
What it tests
- Easy-to-medium coding problems on strings, arrays and trees
- Multiple-choice questions on ML basics and LLMs
- A short applied ML task in some versions
- Working steadily through a long, timed test
How to prepare
- Practise timed sets of easy and medium problems in a browser editor.
- Revise ML and LLM vocabulary for the multiple-choice part: overfitting, metrics, ensembles, transformers, fine-tuning.
5 candidate reports, 4 from 2024-2025 intern, ML assistant and AI engineer roles plus one older data scientist report; experienced-hire reports do not mention it.
03Technical screen
About 60 min video call: short-answer ML questions and resume discussion plus a coding task; one older applied scientist process had two phone screens
The gate before the loop for applied scientist and ML engineer roles, and in some data scientist processes. Interviewers ask quick ML questions (overfitting, underfitting, gradient descent, LLM topics), dig into your resume and give a coding problem, sometimes on HackerRank. One candidate described a screening call that turned deeply technical with little warning.
What it tests
- Fast, accurate answers to core ML questions
- Explaining your recent projects and the choices behind them
- Solving an easy-to-medium coding problem while talking through it
- Relating your experience to the team's AI work
How to prepare
- Prepare two-sentence answers to standard ML questions; the screen moves quickly.
- Treat any screen as technical even if the invite sounds casual.
- Practise coding in a plain editor and narrate your approach before you write code.
10 candidate reports describe a technical screen before the loop; 6 are applied scientist or ML engineer roles since 2024, consistently ML questions plus coding in about an hour.
04Hiring manager conversation
30-60 min with the hiring manager, a director or a VP; early in most data scientist processes, at the end of some loops
Data scientist candidates usually meet the hiring manager right after the recruiter, and the talk ranges from a description of the team's work to a technical review of your projects. In other processes a manager or director round closes the loop. One candidate expected a casual chat and instead met an unannounced team lead with technical questions, so prepare as for a technical round.
What it tests
- How your past projects map to the team's problems
- Justifying the approach you chose on a project and its results
- Motivation, career goals and fit with the team
- Technical depth in the team's area, for example computer vision or LLMs
How to prepare
- Prepare a five-minute and a fifteen-minute walkthrough of your strongest project, including what you would change.
- Ask the recruiter who you will meet and whether the round is technical.
- Use the manager's description of the team to steer later answers toward its problems.
16 candidate reports (9 since 2024) include a hiring manager, director or VP round; data scientist reports usually place it second, other tracks at the end of the loop.
05Coding rounds
One to three 45-60 min rounds in the loop, live in an online editor such as HackerRank or a shared pad; Python, SQL or algorithms depending on role
Oracle's coding is mostly practical. Data scientist rounds pair Python (functions, iteration, object-oriented code, numpy or pandas) with SQL joins, including self joins across several tables. Applied scientist and ML engineer loops add one to three algorithm rounds, typically easy to medium: string rotation, anagrams, tree height or comparison, linked lists, and in one case refactoring code. Interviewers in several reports cared more about reasoning than exact syntax.
What it tests
- Easy-to-medium algorithms on strings, trees and linked lists
- SQL with self joins and multi-table joins
- Clean Python with functions, classes and object-oriented design
- Parsing and recursion, for example turning an expression into a syntax tree
- Refactoring or extending working code
How to prepare
- Drill SQL self joins and multi-table joins; they are the most specific SQL detail in our reports.
- Be ready to explain object-oriented ideas with a small Python example.
- Oracle's careers blog advises clarifying vague problems, outlining the plan before coding and walking through a test case; an unfinished but well-explained solution can still score.
- Use the language you know best; Oracle says any language is fine unless the interviewer specifies one.
27 candidate reports (14 since 2024) describe live coding; difficulty is mostly easy to medium. Oracle's careers blog describes live coding interviews.
06ML, statistics and LLM fundamentals
One to three 45-60 min discussion rounds, sometimes half of a combined ML and coding interview; mostly verbal, little code
The most reported content. Interviewers check classic ML (supervised vs unsupervised, regression, SVM, decision trees, random forests, bagging vs boosting, k-NN), handling of messy data (imbalance, outliers, sparse data, cold start) and statistics (central limit theorem, normality, stationarity). Deep learning questions cover vanishing gradients, LSTMs and embeddings. Since 2023 applied scientist and data scientist rounds increasingly ask about LLMs: fine-tuning, LoRA and PEFT, transformer vs LLM architecture and RAG. Some older data science reports used a feature-engineering case on a few tables of web or shopping data.
What it tests
- Ensembles: how random forests and boosting work and why they reduce error
- Overfitting, underfitting, gradient descent and vanishing gradients
- Imbalanced data, outliers, sparse features and cold start
- Statistics basics: central limit theorem, normality checks, stationarity
- LLM fine-tuning (LoRA, PEFT), transformer architecture and RAG
How to prepare
- Be able to explain random forests and bagging vs boosting crisply; tree ensembles are the most repeated classic ML topic.
- For OCI and other AI teams, prepare LLM fine-tuning and RAG answers with real numbers from your own work.
- Expect follow-up questions on any method named in your projects, for example why you picked an unsupervised segmentation technique.
35 candidate reports (about 13 since 2024) describe ML, DL, statistics or LLM questions; LLM topics appear in most applied scientist reports since 2023.
07System and ML system design
45-60 min design discussion in the loop; reported mainly for applied scientist, ML engineer and MLOps roles
Applied scientist and ML engineer loops often include one design round. Prompts in our reports range from an ML system (a recommendation system with sparse data and cold start, an ML pipeline, a RAG system) to general backend design (a ride-hailing rider system, distributed log processing). One candidate summed it up as designing a full system from start to end, close to real applications.
What it tests
- Framing an ML problem end to end: data, model, evaluation, serving
- Recommendation design with sparse data and cold-start users or items
- Retrieval-augmented generation pipeline design
- Scalable backend components: APIs, storage, load balancing, distributed processing
How to prepare
- Practise one recommendation design and one RAG design until you can cover data, model, metrics and serving in 40 minutes.
- For MLOps and senior ML engineer roles, also prepare a classic backend design, since some prompts had no ML in them.
7 candidate reports (5 since 2024), mostly applied scientist, ML engineer and MLOps roles; prompts split between ML or RAG design and general backend design.
08Project presentation or take-home (some data science teams)
Presentation of a past project or of a take-home data challenge to the team, followed by Q&A
Some data scientist loops ask you to present your own ML projects in the final stage, and a few teams instead give a take-home data challenge (one candidate had a week) whose results you present. One UK take-home asked for analysis, modelling, a presentation and unit tests on all the code, which the candidate found long alongside a full-time job.
What it tests
- Explaining a project's problem, approach, results and limits to a mixed audience
- Handling detailed questions on modelling choices
- For take-homes: clean analysis, modelling and tested code
How to prepare
- Prepare a 15-minute project talk with clear slides on data, method, results and what you would do next.
- If you get a take-home, ask how long it should take and budget time for tests and slides, not only the model.
5 data scientist candidate reports: three project presentations (two since 2025) and two take-home challenges with a presentation (2019 and 2023).
09Behavioral and culture rounds
One or two 45-60 min behavioral interviews in the final loop, or behavioral questions inside technical rounds
Most loops include behavioral questions, either as their own rounds or mixed into technical and manager interviews. Oracle's careers blog says behavioral questions come in the final interview phase, uses the STAR format and points candidates to the company's core values in its interview prep packet. Candidates report questions on conflict, unhappy clients, setbacks and why they want to move.
What it tests
- STAR stories on conflict, ambiguity and difficult choices
- Handling an unhappy client or stakeholder
- What went wrong on a project and what you learned
- Fit with Oracle's values and the team's way of working
How to prepare
- Prepare five or six STAR stories and map each to a value in the prep packet your recruiter sends.
- Keep the reason for leaving your current job positive, as Oracle's own guidance advises.
- Expect a conflict scenario that escalates through follow-up questions.
11 candidate reports (5 since 2024) describe behavioral rounds or questions in the loop. Oracle's careers blog describes behavioral questions in the final phase, using STAR.
Questions and themes reported most often
- Walk through your projects and justify the approach you chose (12 reports, hiring manager conversation)
- Present a past ML project to the team, then answer questions (5 reports, project presentation or take-home (some data science teams))
- How random forests, bagging and boosting work (4 reports, ml, statistics and llm fundamentals)
- Easy algorithm problems: string rotation, anagrams, tree height or identical trees, linked lists (4 reports, coding rounds)
- LLM fine-tuning (LoRA, PEFT) and transformer vs LLM architecture (4 reports, ml, statistics and llm fundamentals)
- Overfitting, underfitting, gradient descent, vanishing gradients and imbalanced data (3 reports, ml, statistics and llm fundamentals)
- Engineer features from a few tables of web or shopping data (3 reports, ml, statistics and llm fundamentals)
- Design a full system end to end: ride-hailing, log processing (3 reports, system and ml system design)
What Oracle asks about
- ML, DL & LLM concepts32%
- Coding (Python & algorithms)22%
- Behavioral & manager16%
- ML / system design12%
- SQL & data10%
- Projects & presentation8%
Oracle interview prep checklist
- Solved easy-to-medium string, tree and linked list problems while explaining the plan first
- Wrote SQL self joins and multi-table joins and explained an OOP concept with Python code
- Can explain random forests, bagging vs boosting, overfitting and vanishing gradients in two minutes each
- Can discuss LoRA or PEFT fine-tuning, transformer architecture and a RAG design
- Sketched a recommendation system with cold start and sparse data, end to end
- Prepared a project presentation and a short version of the same story
- Have five STAR stories mapped to Oracle's values and confirmed the role level with the recruiter
Tips
- Ask the recruiter for the type and length of every round and whether a presentation is expected; Oracle's loop varies by team.
- Confirm the role level before the loop; a senior ML candidate reported learning afterwards that the opening was junior.
- Split preparation between practical coding (Python, SQL joins, easy-to-medium algorithms) and ML fundamentals; that mix covers most rounds.
- Know tree ensembles, overfitting, imbalance and outlier handling cold; these classic ML questions recur across a decade of reports.
- For OCI AI and other applied scientist roles, prepare LLM fine-tuning, transformer and RAG answers plus one end-to-end system design.
- Rehearse a project talk in two lengths; data science loops often ask you to present your own work.
- Prepare STAR stories mapped to the core values in the interview prep packet.
- Treat every call as possibly technical; candidates met unannounced interviewers and screens that went deep.
Practice the Oracle loop
Work the stages in order: every reported question for this company, grouped by round, with solutions in the app.
Oracle interview FAQ
How many interview rounds does Oracle have for ML roles?
Candidates report 9 stages: Recruiter or HR screen, Online assessment (interns and early career), Technical screen, Hiring manager conversation, Coding rounds, ML, statistics and LLM fundamentals, System and ML system design, Project presentation or take-home (some data science teams) and Behavioral and culture rounds. A recruiter or HR screen comes first in most reports. Applied scientist and ML engineer candidates (often OCI AI teams) then have a technical screen of about an hour that mixes short ML questions, resume discussion and a coding task, followed by a loop of four or five interviews (one report says about four hours) covering coding, ML or AI fundamentals, system design and behavioral questions. Data scientist candidates usually meet the hiring manager early, then face technical rounds on ML, statistics, Python and SQL, and a final stage of four to six interviews that may include a project presentation. Interns and some early-career roles take an online HackerRank test first. A few data science teams have used a take-home data challenge. The outline rests on short candidate reports; only about eight describe an OCI-style applied scientist or ML engineer loop since 2024.
How long does the Oracle ML interview process take?
Reported timelines run 2-8 weeks (Usually a few weeks; one report waited months).
What does Oracle ask in ML interviews?
By reported question mix: ML, DL & LLM concepts 32%; Coding (Python & algorithms) 22%; Behavioral & manager 16%; ML / system design 12%; SQL & data 10%; Projects & presentation 8%. The most repeated themes: Walk through your projects and justify the approach you chose; Present a past ML project to the team, then answer questions; How random forests, bagging and boosting work.
How should I prepare for the Oracle ML interview?
Ask the recruiter for the type and length of every round and whether a presentation is expected; Oracle's loop varies by team. Confirm the role level before the loop; a senior ML candidate reported learning afterwards that the opening was junior. Split preparation between practical coding (Python, SQL joins, easy-to-medium algorithms) and ML fundamentals; that mix covers most rounds.
Are there free Oracle interview questions?
LeetML has 2 questions reported from Oracle 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
- IBMA HackerRank coding and SQL test, then resume-driven ML and GenAI interviews and a manager round; IBM Research runs a research talk instead.
- MicrosoftProject deep dives in most rounds, quick ML and LLM fundamentals, standard coding, from-scratch ML code and GenAI design, in a loop shaped by the hiring team.
- AmazonLeadership Principle questions in every round, standard algorithm coding, and ML breadth, depth and design rounds that differ by track.
- SalesforceA team-run loop that mixes plain coding, ML and LLM fundamentals, applied design and a deep look at your own work.
- SAPAn online coding test, an HR screen, then project-driven technical rounds with easy-to-medium coding and a hiring-manager conversation.
- DatabricksPractical coding, LLM-centred ML design on data-platform problems, and a long, structured panel.