What does your current schema look like? If you have a field with a type of something like ["null","int"] then it is being declared as a "nullable union", meaning the value can be null or a valid integer. If instead you use simply "int" for the type, then it should enforce non-null values for that field. If it does not, then the CSV reader from the Kite SDK (used to parse the CSV in the ConvertCSVtoAvro processor) likely treats missing values as empty or default rather than null. If this is the behavior you're seeing, please feel free to file a Jira to improve the handling of missing CSV values.