Goldman Sachs Data Scientist Interview Questions

1.Tell me about someone junior you've mentored on data science or analytics skills. What specific skill did you teach them, and how did you measure their progress?

2.Tell me about a time when you realized your approach to a data science problem was fundamentally flawed. What made you recognize this, and how did you pivot?

3.Tell me about a time when you had to work with both quantitative researchers and engineers to deliver a critical model or analytics product. What was your role in bridging those groups?

4.Walk me through a time when you had to present negative findings or disappointing model results to business stakeholders. How did you frame the message and what was their reaction?

5.Describe a time when a model or analysis you built performed poorly in production or failed to meet expectations. Walk me through what went wrong and how you applied those lessons to a subsequent project.

6.Describe a project where you needed input from risk management, trading, and technology teams to build an analytics solution. How did you coordinate across these different groups?

7.Implement a function to calculate the maximum drawdown of a portfolio given a time series of daily returns. What's the time and space complexity?

8.Write a function to merge two sorted arrays of stock prices, where each array represents prices at different timestamps. The result should be sorted by timestamp.

9.Walk me through how you'd design an A/B testing framework for the Marcus savings product. We need to test interest rate changes while ensuring we don't violate regulatory requirements.

10.You have a dataset of trade executions stored in kdb+ that's 500GB. A portfolio manager wants daily P&L attribution analysis, but the current query takes 45 minutes. How do you make this faster?

11.You're building a risk model that needs to process real-time market data feeds for thousands of securities. The model currently takes 200ms to update, but traders need results within 50ms. How would you optimize this?

12.You're tasked with building a fraud detection model for the Marcus credit card product. Explain your approach to handling the severe class imbalance where fraud represents less than 0.1% of transactions.

13.Design a recommendation system for the Marcus financial planning tools that suggests personalized savings goals and investment strategies. The system serves 5 million active users with diverse financial profiles.

14.Design a data lineage and quality monitoring system for Goldman's data lake that tracks data flow from external sources through transformation pipelines to client-facing applications like Marquee.

15.Design an anomaly detection system for Marcus credit card transactions that can flag potential fraud in real-time. The system processes 50,000 transactions per minute and needs to respond within 100ms.

16.Design a data warehouse solution for Marquee that allows institutional clients to query 10 years of historical market data and research. Clients expect sub-second response times for complex analytical queries.

17.Design a data pipeline that ingests real-time market data from multiple exchanges and feeds it to our trading algorithms. The system needs to handle 10 million price updates per second with sub-millisecond latency.

18.Design a real-time dashboard system for GS Risk Systems that aggregates position data across all trading desks and displays firm-wide risk metrics. The system must update within 5 seconds of any position change.

19.Tell me about a time you identified a gap in your team's technical skills that was critical for an upcoming project. How did you address it, and what role did you take in developing that capability?

20.Tell me about a time when you had to deliver disappointing results or admit you were wrong about a data science approach in front of senior stakeholders. How did you handle it?

21.Tell me about a time you had to make a critical decision about data quality or model accuracy when stakeholders were pressuring you to move faster. What framework did you use to balance speed against getting it right?

22.Walk me through a time when you had to coordinate data science work across multiple teams that had competing priorities or different definitions of success. What was your role, and how did you align everyone?

23.Describe a situation where you needed to influence a senior trader or portfolio manager to adopt a new analytical approach or change how they used your models. How did you build credibility with someone who had decades of market experience?

24.Describe a situation where you had to advocate for additional resources or budget for a data science initiative when leadership was focused on cost reduction. How did you make the business case?

25.Marcus just launched a new high-yield savings account with a 4.5% APY. Estimate how many new customers we'd need to acquire to cover the annual cost of this rate increase if we raised it from our previous 3.8% rate.

26.If trading volumes across all asset classes dropped 40% next quarter due to market uncertainty, how would you prioritize which technology and data science projects to continue funding versus which to pause?

27.Our Marquee platform shows that institutional clients are accessing research reports 30% less frequently this quarter. Walk me through how you'd investigate whether this reflects changing client behavior or a platform issue.

28.Estimate the potential revenue impact if we could reduce trade settlement time from T+2 to T+1 across our equity trading business. What data would you need and how would you structure this analysis?

29.You notice that our credit risk models for corporate lending are showing higher default probabilities for mid-market companies in the energy sector, but our competitors seem to be pricing more aggressively. How would you investigate whether our models are too conservative or if competitors are mispricing risk?

30.Goldman Sachs is considering launching a digital wealth management platform for mass affluent clients. Estimate the potential market size and what client acquisition cost would make this profitable, given our current cost of capital and technology investments.