Autocoder

BNY interview questions with answers (technology roles)

Last updated 2026-11-19

BNY hires technology graduates and early-career engineers in India. Candidates commonly describe a process with an online assessment followed by technical and HR interviews. Technical questions tend to centre on Java, Spring, SQL and Python, with practical topics such as batch jobs, testing and version control, and some interest in the custody and settlement business the bank is known for.

The thirteen questions below are practice in that style, with worked answers. They are not past questions. For the online assessment, see BNY coding round questions.

Typical rolesGraduate software engineer, technology associate (India)
Rounds, as commonly reportedOnline assessment, technical interview, HR round
Heavy topicsJava, Spring and JPA, SQL, Python, batch jobs, testing
Hiring newsSee Hiring status below

Rounds as commonly reported

Candidates commonly report an online assessment with coding and technical questions, followed by a technical interview and an HR discussion. The format varies by programme and year, so use your own invitation as the source.

In the technical interview, expect questions to follow your resume, with practical scenarios such as designing a restartable batch job or explaining why a query is slow. Domain questions are usually light, but a short explanation of custody and settlement shows interest.

What to prepare

Hiring status

There is no BNY hiring-news page on this site, and this page quotes no hiring numbers, dates or packages because none are sourced here. Check BNY's own careers site for current India graduate and technology roles, eligibility and dates, and apply only there.

BNY interview questions with answers (technology roles): questions with worked answers

  1. 1

    What are Java generics and type erasure, and what limits does erasure impose?

    medium

    Generics let a class or method work with a type parameter checked at compile time, so List<String> rejects an Integer and removes casts. The compiler erases type parameters to their bounds in the bytecode, so at run time a List<String> and a List<Integer> are the same class. As a result you cannot create new T(), new T[n], use instanceof List<String> or overload methods that differ only in type arguments. Wildcards express variance: ? extends T for reading (producer) and ? super T for writing (consumer), the PECS rule.

  2. 2

    Explain map versus flatMap in Java streams and build a map of customer to total amount.

    medium

    map transforms each element into exactly one output; flatMap transforms each element into a stream and flattens them, for example turning a list of orders into one stream of order lines. To total amounts per customer, use Collectors.groupingBy with summingDouble, or toMap with a merge function; toMap without a merge function throws IllegalStateException on duplicate keys. Avoid side effects inside stream operations. For money, prefer BigDecimal with a reduce.

    java
    import java.util.*;
    import java.util.stream.*;
    record Payment(String customer, double amount) {}
    class Demo {
        static Map<String, Double> totals(List<Payment> ps) {
            return ps.stream().collect(
                Collectors.groupingBy(Payment::customer, Collectors.summingDouble(Payment::amount)));
        }
    }
  3. 3

    What is the N+1 select problem in JPA or Hibernate, and how do you fix it?

    hard

    Loading a list of N parents and then touching a lazy collection on each one triggers one query for the list plus N queries for the children. It is invisible in small test data and slow in production. Fixes include a JOIN FETCH in the query, an entity graph, batch fetching (hibernate.default_batch_fetch_size) or a projection that selects just the needed columns. Turn on SQL logging in tests to catch it. Eager fetching everywhere is not a fix, since it loads data you may not need.

  4. 4

    What is the difference between a view and a materialised view, and when do you use each?

    medium

    A view is a stored query that runs each time it is read, so it is always current but as slow as the query. A materialised view stores the query result physically and must be refreshed, on demand or on a schedule, so reads are fast but the data can be stale. Use views for abstraction and security (exposing only some columns) and materialised views for expensive aggregates that are read often and tolerate some delay, such as end-of-day position summaries.

  5. 5

    SQL: find the missing numbers in a sequence of invoice numbers.

    medium

    Compare each invoice number with the next one using LEAD; wherever the next minus current is greater than one, the numbers in between are missing. This reports ranges of gaps, which is usually enough and avoids generating a number series. In PostgreSQL, generate_series with a left join also lists each missing number individually. State your assumptions about duplicates and the first and last numbers.

    sql
    SELECT invoice_no + 1 AS gap_start, next_no - 1 AS gap_end
    FROM (
      SELECT invoice_no, LEAD(invoice_no) OVER (ORDER BY invoice_no) AS next_no
      FROM Invoices
    ) t
    WHERE next_no - invoice_no > 1;
  6. 6

    With pandas, how would you find the three most traded securities per day by volume?

    medium

    Group by date and security, sum the volume, then rank within each date and keep ranks up to three. groupby followed by sum gives one row per date and security; sort_values and groupby.head(3) take the top three per date. This mirrors the SQL window-function approach. For millions of rows, check memory use and consider categorical dtypes or doing the aggregation in the database.

    python
    import pandas as pd
    
    def top3_per_day(df):
        daily = df.groupby(["date", "security"], as_index=False)["volume"].sum()
        return (daily.sort_values(["date", "volume"], ascending=[True, False])
                     .groupby("date").head(3))
  7. 7

    Design an end-of-day batch job that settles transactions. How do you make it restartable and safe?

    hard

    Split the work into steps with a persisted status for each (for example per file or per batch of transaction ids), so after a failure the job resumes from the first incomplete step instead of starting over. Make each step idempotent: use unique keys or upserts so reprocessing a record does not duplicate its effect. Process in chunks inside transactions, validate counts and totals against the input (reconciliation) and write a clear audit log. Alert on failure and on the job not finishing by its deadline. Spring Batch provides chunking and restart metadata.

  8. 8

    What is the role of Mockito in unit testing, and what is TDD?

    easy

    Mockito creates mock objects for a class's dependencies so you test the class alone: you stub return values, then verify that expected calls were made. Mock external boundaries such as repositories and HTTP clients, not the class under test or simple value objects. Test-driven development means writing a failing test first, writing just enough code to pass it, then refactoring, which keeps design simple and gives a regression suite. Over-mocking leads to tests that mirror the implementation and break on every refactor.

  9. 9

    Compare GitFlow and trunk-based development.

    easy

    GitFlow uses long-lived develop and release branches plus feature and hotfix branches, which suits scheduled releases but causes painful merges when branches live long. Trunk-based development has everyone merging small changes into one main branch frequently, protected by automated tests and feature flags, which supports continuous delivery. Many banks that release on a fixed cycle use a hybrid. Whichever you pick, keep branches short-lived and rebase or merge often.

  10. 10

    In one minute, explain what a custody bank does and what trade settlement means.

    medium

    A custody bank safeguards clients' securities and cash, records ownership, collects dividends and interest, and processes corporate actions such as splits. Settlement is the step in which a trade is completed: the buyer's cash is exchanged for the seller's securities, a set number of days after the trade date (for example T+1, one business day later). Problems arise when instructions do not match or securities are not available, which is why reconciliation and accuracy matter so much in this part of the industry.

  11. 11

    Why BNY, and why a technology role in a financial services company?

    easy

    Say what interests you: infrastructure that financial markets rely on, where accuracy and reliability matter more than novelty, and a chance to work on large data and automation problems. Connect it to a project, for example one where you handled data validation. Keep claims about the company to what you have read on its own website, and finish with what you would like to learn in the first year.

  12. 12

    This work can involve detail-heavy checks. How do you maintain accuracy?

    medium

    Describe the methods you actually use: checklists, automated tests and validations, reviewing your own work after a break, pair review and keeping notes on past mistakes. Give an example in which a check caught an error, with the consequence it would have had. Show that you see accuracy as a design problem to automate, not only a matter of being careful.

Frequently asked questions

Which topics does BNY emphasise in technical interviews?

Candidates commonly report Java, Spring, SQL and practical engineering topics. Prepare whatever your resume lists, to the level of explaining why.

Do I need to know about custody banking?

It is not required, but a short, correct explanation of custody and settlement helps show motivation.

Is Python tested?

It depends on the team. Data-oriented roles may involve Python and pandas, so prepare the basics if they are on your resume.

Where can I see current openings?

Use BNY's own careers site. This page quotes no hiring numbers because none are sourced here.

Related guides

More in Company Interview Questions

See all in this section