Goldman Sachs DevOps / SRE Interview Questions

1.Describe a time when a system or process you built failed in production. How did you handle it, and what did you learn that changed how you approach similar problems?

2.Tell me about a time you implemented a new monitoring or automation solution that initially didn't work as expected. How did you turn it around?

3.Tell me about a time you had to coordinate with multiple engineering teams to resolve a production outage that was impacting client-facing systems. Walk me through how you managed the response.

4.Tell me about a time you coached a junior engineer through a complex production issue or infrastructure challenge. What was your approach, and how did it turn out?

5.Walk me through a time you had to tell a business stakeholder that a system they depended on would be down longer than expected. How did you handle that conversation?

6.Describe the most complex cross-team project you've led where you had to align engineering, operations, and business teams on infrastructure changes. What was your approach to getting everyone on the same page?

7.You're monitoring our Marquee platform and notice that API response times for portfolio analytics are degrading during market open. CPU and memory look normal, but P95 latency jumped from 50ms to 800ms. How would you diagnose this?

8.Write a Python script that monitors Kafka lag across multiple consumer groups and alerts when any partition lag exceeds a threshold. The script should handle authentication and run in our Kubernetes environment.

9.Our kdb+ database storing tick-by-tick market data is running out of disk space faster than expected. The data retention policy says 2 years, but we're filling up after 18 months. What's your approach?

10.You need to deploy a configuration change across 500 microservices in our GS Engineering Platform, but some services are more critical than others. Design a rollout strategy that minimizes risk to trading systems.

11.We're migrating Marcus consumer banking workloads from on-premises to AWS, but the compliance team is concerned about PCI DSS requirements for payment processing. How do you design the infrastructure to meet both cloud scalability and regulatory needs?

12.Our Transaction Banking platform processes millions of payment messages daily across multiple regions. Design an auto-scaling strategy for the Kubernetes pods that considers both message volume and regulatory latency requirements.

13.Design a log aggregation and search system for Marquee that can handle 500GB of application logs daily while allowing portfolio managers to troubleshoot client-specific issues quickly.

14.We're implementing chaos engineering for our Transaction Banking platform. How do you design failure injection that tests resilience without risking real payment processing or violating regulatory requirements?

15.We're building a real-time fraud detection pipeline for Marcus that needs to evaluate every transaction within 100ms while handling 50,000 transactions per second during peak hours. How would you architect this?

16.We need a secrets management system for our Kubernetes clusters that handles API keys, database credentials, and trading system certificates. How do you design this to meet both security and developer productivity needs?

17.Design a circuit breaker system for our GS Risk Systems that needs to handle millions of position updates per second. When should it trip, and how do you prevent cascading failures across dependent risk calculations?

18.Design a deployment orchestration system for the GS Engineering Platform that can safely deploy code to 2,000+ microservices while ensuring trading systems never experience downtime during market hours.

19.Tell me about a time you had to convince a senior engineer or architect to change their approach to an infrastructure design. How did you build that case?

20.Tell me about a time you had to learn a new technology or system quickly to solve a critical problem, then teach that knowledge to your team. How did you approach both the learning and teaching?

21.Describe a time you had to make a quick infrastructure decision during a production incident that could have impacted trading systems. Walk me through your decision-making process.

22.Tell me about a time you had to mentor someone on your team through a critical failure in production. How did you balance teaching with getting systems back online?

23.Our Marquee platform team and Risk Systems team are both claiming that infrastructure changes the other team made are causing performance issues. As a DevOps engineer, how do you approach resolving this cross-team conflict?

24.Describe a time when you had to convince business stakeholders to invest in infrastructure improvements that wouldn't deliver immediate visible features. What was your approach?

25.Our Marcus banking platform had 99.95% uptime last month, but we're targeting 99.99%. That gap costs us about $2M annually in SLA penalties and customer churn. Walk me through how you'd estimate the infrastructure investment needed to close that gap.

26.You're investigating why our GS Engineering Platform's deployment success rate dropped from 98% to 94% over the past quarter, but no major changes were shipped. How would you approach this investigation?

27.Estimate the networking bandwidth cost if we moved all Marquee platform traffic from our on-premises data centers to AWS. The platform serves 50TB of market data daily to institutional clients. What are your key assumptions?

28.Our Risk Systems are processing 500M position updates per hour during peak trading, but we're seeing intermittent 5-second delays in risk calculations. The business says any delay over 1 second creates regulatory exposure. How do you systematically identify the bottleneck?

29.Goldman's Transaction Banking platform processes $500B in payments annually across 40 countries. If we wanted to reduce our infrastructure costs by 30% while maintaining regulatory compliance, how would you approach this optimization?

30.We're planning to implement automated failover for our kdb+ market data systems that serve live trading. The business wants 99.99% uptime, but traders are worried about data inconsistencies during failovers. How do you balance these competing requirements?