Software Engineer Interview Questions

At my previous company, our team was considering using a monolithic architecture for a new service I believed should be microservices-based for scalability. I prepared a 2-page technical comparison, presented both options to the lead architect, and suggested a spike to validate the microservices approach. After the spike confirmed better scaling characteristics, the team adopted my recommendation. Key lesson: back disagreements with evidence, not opinions.

During a critical API migration, our team was 2 weeks from launch when we discovered a performance regression. I organised a war room, split the debugging tasks across 3 engineers, and personally stayed late to rewrite the caching layer. We hit the deadline with 4 hours to spare. I logged the root cause in our post-mortem system to prevent recurrence.

I advocate for a 70/30 split: 70% feature work, 30% tech debt reduction in each sprint. I track tech debt in a shared board and pitch debt items by attaching them to business outcomes — e.g. 'fixing this database index will reduce our P95 latency by 40ms, directly improving our checkout conversion.' This framing gets buy-in from product managers.

Vertical scaling adds more resources to a single machine (more RAM, CPU). Horizontal scaling adds more machines. Use vertical for stateful services where distributing state is complex, and horizontal for stateless services where you need high availability and can distribute load. Most modern cloud architectures prefer horizontal scaling for resilience.

Core components: (1) POST endpoint to accept long URL, generate short code (Base62 encoding of an auto-incrementing ID or MD5 hash), store mapping in Redis for fast reads + PostgreSQL for persistence. (2) GET endpoint to look up short code, return 301 redirect. Scaling: CDN for caching hot redirects, read replicas for DB. Rate limiting via Redis sorted sets. Analytics: log clicks asynchronously to Kafka.

1) Alert triage: check PagerDuty severity, confirm blast radius (how many users affected). 2) Mitigation first: rollback last deployment via feature flag or deploy revert — don't debug in prod. 3) Communication: post in #incidents Slack channel, update status page within 5 minutes. 4) Root cause: check logs (Datadog/Splunk), recent deployments, DB error rates. 5) Post-mortem within 48 hours with timeline, root cause, and preventive actions.