Sanjaay Gowda T S
All company profiles

COMPANY CONTEXT / 54 IN THE STUDY LIST

Morgan Stanley

FinancialsUnited StatesSystem designAI & machine learning

Morgan Stanley provides investment banking, wealth management and investment-management services.

Original company-context practice, not verified company interview questions. Use the business setting to explain technical choices clearly.

01 / PRACTICE IN CONTEXT

Two problems to reason through.

Clarify assumptions before reading the model answer. Discuss what would change with different constraints.

Original company-context practiceHard

System design

Produce a consistent hypothetical portfolio view

Morgan Stanley · Concept

Original practice scenario: A simulated wealth dashboard combines positions, prices and foreign-exchange rates that refresh independently. How should a snapshot communicate consistency?

Read model answer

A portfolio total is meaningful only with a defined valuation time and completeness policy. Independently latest values can form a combination that never existed as a coherent snapshot.

Work through it

  1. Version position events and market observations with source and effective times; define acceptable freshness and missing-price behavior before calculating totals.
  2. Build an immutable valuation snapshot identifying the exact position, price and currency-rate versions, with explicit stale or unavailable components.
  3. Test delayed feeds, price corrections, corporate-action boundaries and partially unavailable currencies; expose snapshot status alongside its timestamp.

Tradeoffs to discuss

  • Waiting for aligned inputs improves comparability but increases latency; provisional snapshots are faster if their limitations remain visible.

Common pitfalls

  • Do not replace a missing price with zero without a declared policy.
  • Keep the exercise focused on data consistency rather than presenting investment recommendations.
Open full question
Original company-context practiceHard

AI & machine learning

Retrieve authorized passages from a synthetic research library

Morgan Stanley · Concept

Original practice scenario: Design a research assistant over fictional analyst documents with different client entitlements and frequent revisions.

Read model answer

Authorization belongs in retrieval and evidence handling, not merely in the final prompt. A fluent answer is unacceptable if it uses a passage the requesting user cannot access.

Work through it

  1. Attach document version, publication time and access metadata during ingestion; enforce the user's current entitlement before returning candidate chunks.
  2. Evaluate retrieval and citation support with synthetic questions, conflicting versions and no-answer cases; prefer current authorized evidence under a documented policy.
  3. Recheck permissions when opening cited material, record access-safe diagnostics and test revoked access, malicious document instructions and expired sessions.

Tradeoffs to discuss

  • Filtering before search protects boundaries but can reduce recall; overfetching must still prevent unauthorized text from reaching the model or logs.

Common pitfalls

  • Do not assume a citation proves that every generated sentence is supported.
  • Never place real client documents into the practice dataset.
Open full question

02 / STRENGTHEN THE FOUNDATIONS

Related technical practice.

Browse company questions

03 / TEST AN APPROACH

Suggested coding exercises.

Exercises selected for these technical tracks. They are independent practice with visible tests and public solutions.

Study list: PwC Global Top 100 companies by market capitalisation, 2026 · 31 March 2026.

Keep learning beyond the interview.

Explore the Learning Hub