System Design Interview Questions

Kai W. - "The user awakes Alexa by saying the "Alexa word" Device starts recording audio from the user, and streaming it to the Alexa Cloud Service The recorded audio stream is sent to the ASR (Automatic Speech Recognition) system ASR returns the words and the order of the words in the audio stream, and sends this information to NLU (Natural Language Understanding) NLU returns the intent of the audio stream, and sends it to Alexa Skill If the intent triggers a "play" directive (e.g. turn"See full answer

Sri A. - "Mission: to help users buy and sell products from groups. If we think about the value here, users want to join groups that sell/buy products of interest (young moms or vinyl), or location of interest, or trust, or some other factor - either from other users or SMBs. User Actions: create groups, join groups, active in groups - make posts to sell or show interest in buying (like, commenting, sharing or ideally transacting) deeper engagement - sharing , saving , searching more based on wh"See full answer

Luca D. - "The real discussion was very much similar o what exposed at https://www.tryexponent.com/courses/software-engineering/system-design/design-rate-limiter, but - as I commented the video - the real interviewer wasn't so naive to do not forgive the client identification only because IP. I had to introduce glimpses of https://en.wikipedia.org/wiki/Knowyourcustomer practice, I quoted JWT. I proposed a logical map of id addressing a "deque" of time-stamps of requests, with a threshold for the deque an"See full answer

Clayton P. - "I've watched all the ML Systems designs interviews and this solution provides a clean baseline for predicting ETA using historical averages, but it falls short of addressing the broader problem of route planning. The system predicts ETA for a given segment and time interval, but it doesn’t explain how to compute the ETA for an entire route or how to integrate this into dynamic path selection. It also lacks depth on handling real-time data, adapting to distribution shift, or reacting to sudden"See full answer

James H. - "I broke the vehicles down by regions in the world. Designed a wireless api interface to sync with a database backend that linked to google maps through another service and api."See full answer

Vishal B. - "Design an elevator for the blind Clarification Questions A person who is not able to see at all or a person who can partially see? - completely blind Are we we focusing on the people who are visually impaired but, no other physically challenge? - Yes Goal Company already have the elevator products and now company is trying to enter into new market space. Users Blind users with some assistance - we are not focused on these people as they already have some a"See full answer

Ramesh Y. - "Clarification - is it Uber that is building the feature? Answer: Yes Assuming if Uber is the one building this feature, I would start with if and how does the feature help Uber's mission and north star metrics. Uber's mission is to provide reliable transportation around the world and the primary metric that i can think on a high level is Number of rides , Number of drivers and from business revenue perspective revenue per ride, revenue per rider and driver. I picked revenue per ride because it"See full answer

Aditya O. - "Why is it answered like a Product Design Question? Shouldn't this be a strategy type answer analyzing the pros/cons of the service etc.?"See full answer

Anonymous Jellyfish - "Microservices for resolving diffs, storage S3 for keeping files"See full answer

Bini T. - "I understand this is more focused on ML. However, I have a system question. If users allow us to access their location, or they send location via text box, could we use CDNs for the search without hitting our database? We only query the database when we have zero information on location. Other questions: does embedding always guarantee information on location? Do we discharge the user images after we return a prediction? I heard the feedback that we should keep it for future learning. What would"See full answer

Istvan K. - "Thanks you for sharing insight about the app which I would liek to summarize in 2 structures: functional 2 technical requirement. Requirements: Functional requirement: Able to share messages, siles , and other content like photos, sounds Create groups for conversaion Size: 500 million users, 50m DAU -> 500M active chats daily Files not stored in the system but is can be shared.. Integaration with other APPs Technical requirement: Availability: Can you help me to understand the a"See full answer

Yash G. - " import lombok.*; import lombok.extern.slf4j.Slf4j; import org.springframework.boot.CommandLineRunner; import org.springframework.boot.SpringApplication; import org.springframework.boot.autoconfigure.SpringBootApplication; import org.springframework.stereotype.Component; import org.springframework.stereotype.Repository; import org.springframework.stereotype.Service; import java.util.*; import java.util.concurrent.ConcurrentHashMap; import java.util.concurrent.ConcurrentMap; import java."See full answer

Rashmi T. - "It depends on how computers are distributed. Are they located within same network (behind same switch or router) or distributed across globally? Also, it depends upon if there are any security related encryption or decryption algorithms being executed. Similarly, there are so many factors involved in it. However, if a system is highly responsive then this time might be less than 1 second. For example, any google search responds in less than 1 sec. It is an interesting question and I am eager t"See full answer

Abhi R. - "TCP server/client to facilitate high throughput of incoming metrics.(This is the standard used in the industry by Datadog etc). NoSQL or timeseries database for storing metrics. Timeseries databases are better for aggregating metrics in a given time period. REST API for serving metrics data to visualization frontend."See full answer