• Ability to inspect, clean, and manipulate/prepare data
ㅤㅤ• Discover missing values, invalid values, skewed distribution in a dataset that could impact training
ㅤㅤ• Identify relevant transformations and their benefit
• Demonstration of knowledge about using and training common machine learning models
ㅤㅤ• Given a high-level problem, identify the problem type and candidate models
ㅤㅤ• List common models, when each is appropriate, and how it works behind the scenes
ㅤㅤ• Describe tradeoffs involved between different models, along with considerations needed when choosing between them
ㅤㅤ• Describe relevant hyperparameters for a given model and a procedure for selecting optimal values
ㅤㅤ• Describe various "ensemble" methods that could improve model performance
ㅤㅤ• Split a dataset into multiple parts in a way that optimizes model performance
ㅤㅤ• Understand underfitting and overfitting, detection and how to address
• Ability to monitor, evaluate, and maintain deployed ML systems as they evolve
ㅤㅤ• Detect and solve data quality problems like distribution shift, skewed datasets, missing values
ㅤㅤ• Describe suitable metrics to measure model performance and tradeoffs between them
• Ability to efficiently implement ML systems using algorithms that minimize training time, memory usage, or other limited resources
ㅤㅤ• Online/streaming algorithms
ㅤㅤ• Efficient access and usage of available memory
ㅤㅤ• Parallel and distributed training