Machine Learning Engineer Interview Questions

Kartikeya N. - "C : Okay. So I would want to start with knowing what is the product for which we have to build a recommendation system. I : This is a photo sharing product. C : Okay. So is this something on the lines of Instagram? I : Yes C : Okay. And are we a new product co or we have some current product built already? I : You can assume yourself. C : Okay. Is there any demography or country we are targeting? I : No, this is a global product C : Okay. So, the biggest goal of any product recommendation system"See full answer

Jacky Y. - "Problem scope: Can this system detect Bot in real-time online or offline? Both. Online traffic: 1M DAU. Latency: 2s. Offline frequency: daily Offline data: 2B activity logs. Data: How do we know a Bot player (Label)? Human label. Imbalance data: reweight, resample. Develop a Bot simulator to generate more data offline for training. Given lower weight to simulator data than human label. Features: Signals from different models online. Log all the features for offline. Propose new features: dail"See full answer

Alfred O. - "We can use dictionary to store cache items so that our read / write operations will be O(1). Each time we read or update an existing record, we have to ensure the item is moved to the back of the cache. This will allow us to evict the first item in the cache whenever the cache is full and we need to add new records also making our eviction O(1) Instead of normal dictionary, we will use ordered dictionary to store cache items. This will allow us to efficiently move items to back of the cache a"See full answer

Mrinalini R. - "A good answer is describing an experience where you either proposed something that was selected after discussion or an alternate approach was taken, or you disagreed with a decision and argued for something else, either successfully or not. A good answer would be you had cogent arguments, the decision went another way for some reason, and you then fully backed the decision (agree to disagree is another way of stating it). You did not take it personally and you fully committed to the deci"See full answer

VContaineers - " Compare alternate houses i.e for each house starting from the third, calculate the maximum money that can be stolen up to that house by choosing between: Skipping the current house and taking the maximum money stolen up to the previous house. Robbing the current house and adding its value to the maximum money stolen up to the house two steps back. package main import ( "fmt" ) // rob function calculates the maximum money a robber can steal func maxRob(nums []int) int { ln"See full answer

Lati K. - "I generally struggle with stakeholders and partners who doesn't communicate enough. Now it could be either they don't invest sufficient time and energy in doing so or at times they lack the skill sets to do so. In both the cases, the entire responsibility fell on the other person to dig deep into why someone is doing the way they are doing, reading into patterns and behaviour of their personality and adapting to those communication styles"See full answer

Rajesh V. - "As a PM i received a feedback from my program manager on my style of verbal communication. It is about me speaking faster when i wanted to get away with a topic that i wasn't confident (may be not backed up with data, or still in process of getting detailed insight of a problem etc.). Whereas when I'm confident I tend to speak slowly or more assertively that made people to follow easily. I welcomed that feedback so from then on when I'm not confident in a topic I became more assertive to let pe"See full answer

Michael F. - "Prompt: We work for an online shopping website. Our team wants to consider offering discounts (e.g. 10% off your next purchase) to customers to incentivize them to make purchases. How would you design a system that decides how to offer these incentives? Answer Goals: Increase customer engagement while controlling costs. Specifically, we want the increase in revenue per customer per week of customers that receive the discount to be greater than the cost of the discount. Metrics: Revenue per cu"See full answer

Anonymous Prawn - "Would be better to adjust resolution in the video player directly."See full answer

Tian H. - "slack. it's one of the daily productivity tools I need to use. slack is to make the collaborative work quicker, easier, and less formal. it's an instant message app with different features built on. the main user value is to improve the communication efficiency. the main business value is the paid seats and revenue. I like so many things but also suffer from other things. slack target enterprise users, community users, smb users. I fall under the first 2 buckets. personally I have high torre"See full answer

Clayton P. - "The interviewer hinted that a two-tower recommender system might be a suitable approach, using user history to embed users and pages separately and train on view or interaction data. Instead, I proposed a different approach that I felt was more aligned with how knowledge is structured in Confluence: I designed a system using a graph database to model the relationships between Confluence pages. Each page is a node, and edges represent content-based references. For example, when one article"See full answer

Kishor J. - "we can use two pointer + set like maintain i,j and also insert jth character to set like while set size is equal to our window j-i+1 then maximize our answer and increase jth pointer till last index"See full answer

Sai vuppalapati M. - "At a high level, the core challenge here revolves around building an effective recommendation algorithm for news. News is an inherently diverse category, spanning various topics and catering to a wide array of user types and personas, such as adults, business professionals, general readers, or specific cohorts with unique interests. Consequently, developing a single, one-size-fits-all recommendation algorithm is not feasible. To enhance the personalization of the news recommendation algorithm,"See full answer

Gaston B. - "Use a representative of each, e.g. sort the string and add it to the value of a hashmap> where we put all the words that belong to the same anagram together."See full answer