ML system design

• Take 5 minute or so to understand the question

• make assumption by limiting the amount of questions to be asked. just assume any generic numbers

• save time be deliberate with high level system desing digram. instead of xgboost use model and decide which model to use later.

• data consideration -- there are lot of data considration [8-9 min]

  • what are your labels

  • what are your features

  • how to represent data

  • how to split the data

  • how to deal with imbalances

• most impo thing -- talk about

  • modelling

  • metrics

  • training

  • overfitting and other issues

  • cold start

  • time travel problem

• at its code its all about either classification or regression

• last final

  • come up with multiple solution and answer which solution is better and discuss the tradeoffs

Resources --

Evidently AI - ML and LLM system design: 800 case studieswww.evidentlyai.com

A Review of Modern Recommender Systems Using Generative Models (Gen-RecSys)arXiv.org

Recommender Systems - A Complete Guide to Machine Learning Models | Towards Data ScienceTowards Data Science

ML Eng Interview Guide · Patrick HalinaPatrickHalina.com