Start with the role, region, and financial-data context
Morningstar’s technology careers page describes proprietary financial data, improving data quality, and collaboration across research, product, and engineering to build tools for investors.[1] That context motivates the fictional record-processing exercises below; it does not identify an Associate interview syllabus.
Save the job description and note its location, team, required skills, and experience expectations. Check whether your application is for an Associate vacancy, an internship, or a development-program track. Ask your recruiter which technical areas and interview tools apply. Prioritize the posting’s language and framework requirements; the Python example here is a readable practice choice, not a claim about Morningstar’s required language.
For each required skill, prepare one implementation example and one test you can explain. Treat optional technologies as extensions after the fundamentals.
Read public candidate reports without merging their contexts
A public Taro report dated April 1, 2024 concerns an Associate Software Engineer interview in Navi Mumbai, India. Its author mentions basic DSA, development fundamentals, and projects.[2] Use that as one candidate’s preparation context, not evidence of a universal process.
A separate DEV article, published July 24, 2024, is titled “Morning Star SDE-1 Interview Experience (2024).” The author discusses Java, Spring Boot, REST, PostgreSQL, and recent projects; the article does not clearly identify the interview location.[3] SDE-1 is the reported role, not confirmation of the same Associate vacancy. Neither account establishes current US requirements, a fixed round count, or a common question bank.
Independent preparation guide. Not affiliated with or endorsed by Morningstar. Practice questions are original exercises, not an official or verified interview question bank. Interview formats vary by role, team, and location.
Coding: select revised records deterministically
Original exercise: A feed contains repeated price records. Keep the highest revision for each symbol and timestamp, independent of arrival order. Return records sorted by symbol and timestamp. What happens when identical revisions disagree?
Contract: Each tuple is (symbol, timestamp, revision, price_cents). Inputs have validated nonnegative integer revisions and integer cents in one assumed currency; timestamps are normalized UTC strings. Equal symbol, timestamp, and revision must imply equal prices. Conflicting prices raise an error, even for an older revision.
def latest_prices(rows): seen = {} best = {} for symbol, stamp, revision, cents in rows: version = (symbol, stamp, revision) if version in seen and seen[version] != cents: raise ValueError("conflicting revision") seen[version] = cents key = (symbol, stamp) if key not in best or revision > best[key][2]: best[key] = (symbol, stamp, revision, cents) return [best[key] for key in sorted(best)] t = "2026-11-16T10:00:00Z" rows = [ ("AAA", t, 1, 10000), ("BBB", t, 1, 20000), ("AAA", t, 2, 10100), ("AAA", t, 1, 10000), ("AAA", t, 2, 10100), ] print(latest_prices(rows))Expected output:
[('AAA', '2026-11-16T10:00:00Z', 2, 10100), ('BBB', '2026-11-16T10:00:00Z', 1, 20000)]Answer reasoning: A set alone cannot select the winning revision. The seen map detects inconsistent versions; best retains the greatest revision per key. Keeping all versions catches conflicts that a latest-only map could miss. Expected hash-map processing takes O(n) time; sorting k output keys adds O(k log k). Worst-case auxiliary space is O(n).
Follow-up: Test empty input, reversed arrivals, duplicate replays, separate timestamps, and conflicting older revisions. Explain why a rerun should produce the same records or the same rejection.
OOP: separate parsing, selection, and storage
Original exercise: Design a small importer that accepts either CSV or JSON prices without duplicating revision-selection logic.
Answer reasoning: Give each parser a shared parse contract returning validated price records. Pass those records to one selection function, then to a repository that saves the result. Compose the importer from these collaborators rather than inheriting a database class from a parser. Immutable records reduce accidental changes after validation.
Follow-up: How would you test a storage failure? Inject a repository that raises an error and assert that the importer reports failure, not success. Keep parsing tests separate from persistence tests. Add an interface when interchangeable behavior is useful, not merely to increase the class count.
SQL: distinguish missing prices from zero prices
Original exercise: For fictional tables instruments(symbol) and prices(symbol, as_of, price_cents), return instruments without a non-NULL price on 2026-11-16. Assume one price row per symbol and date. Instruments contains AAA, BBB, CCC; prices contains (AAA, 2026-11-16, 10100) and (BBB, 2026-11-16, NULL).
SELECT i.symbol FROM instruments AS i LEFT JOIN prices AS p ON p.symbol = i.symbol AND p.as_of = '2026-11-16' WHERE p.price_cents IS NULL ORDER BY i.symbol;Expected result: BBB, then CCC. BBB has a missing value; CCC has no matching row. AAA is excluded. The date predicate belongs in the join so unmatched instruments remain available for inspection.
Follow-up: Would zero be missing? No: zero is a stored value; its validity needs a separate business rule. Do not replace NULL with zero to hide missing data. Enforce a unique (symbol, as_of) key, or resolve revisions before joining, to prevent duplicate output. This query was checked using local SQLite.
Project discussion: make your contribution inspectable
Original exercise: Explain a project where you transformed input data into something another person could use.
Illustrative answer: “In a coursework CSV importer, I owned row validation and the rejection report. I separated parsing from storage so malformed rows never reached the repository. A three-row test with one missing symbol produced two accepted records and one rejected row with its line number and reason.” This is a fictional answer structure; replace it with work you actually did.
Reasoning and follow-up: Why allow a partial import? In this scenario rows are independent and users can inspect rejected rows. If the file represents one inseparable batch, reject it atomically instead. State which policy your project used, why, and how you tested it. Do not invent performance measurements.
Debugging: investigate a stale API response
Original exercise: A database contains AAA revision 2, but GET /prices/AAA still returns revision 1 after an update. How would you investigate?
Answer reasoning: Reproduce with one symbol and capture the requested timestamp, returned revision, and request identifier. Query the same database and environment used by the service. Trace selection logic before changing infrastructure: an ascending revision sort might select the oldest row. If selection is correct, compare cached and uncached responses. Check whether the cache key omits the timestamp or invalidation failed after commit.
Follow-up: How do you verify a repair? Seed revision 1, request it, commit revision 2, then request again and assert the returned revision and price. Also request another timestamp to expose key collisions. Record observed evidence before declaring the cache guilty.
Rehearse the reasoning, then check the implementation
Choose one exercise that matches your job description. Explain assumptions before coding, walk through the sample, and name an edge case. Run implementation checks in your own local environment. Review whether your explanation identifies the decision rule and why the output follows.
For a separate speaking or writing rehearsal, paste your target job description in the homepage setup, answer by voice or text, and review feedback on your explanation. This is interview-answer practice, not a code runner or an official employer assessment.