Read/Write columns with names known at runtime and store column values into dictionary

[abrasat]

Is it possible to set dynamically at runtime which column names should be read from the csv file and read the values of the selected columns into a dictionary?
As input, I have a list with columns to be read:

public class ColumnNameDescription
{
  public string ColumnName;
  public string ColumnType;
}
 
var dynamicListWithColumnsToBeRead = new List<ColumnNameDescription>
{ 
  {"ColName1", "Int"},
  {"ColName2", "Float"},
  {"ColName3", "Int"}, 
  {"ColName4", "String"}
}

After reading the selected columns, the results should be stored into a dictionary

Dictionary<string, List<object>> csvValuesForColumnsToBeRead = ...

, where the dictionary key is the name of the configured column and the dictionary value is the list with all the values for that column.
Can this be realized with CsvReader?

And also the same question the other way round. I have the dictionary above already containing a list of values to be written for each column:

Dictionary<string, List<object>> csvValuesForColumnsToBeWritten = new Dictionary<string, List<object>>
{
   ["ColName1"] = new List<object> { 1, 7, 55, 12, 155 },
   ["ColName2"] = new List<object> { 12.34f, 88.321f, 456.78f, 100.2f, 888.88f },
   ["ColName3"] = new List<object> { 765, 1234, 4321, 1999, 7 },
   ["ColName4"] = new List<object> { "abc", "def", "ghi", "jkl", "mno" }
}

Can this be written with CsvWriter?
Would be possible to attach different TypeConverters when converting object values to/from string values, depending on the column name ? As there will not be any class with fixed properties, it is not possible to use the ClassMap approach in my scenario. Instead of the class properties it should be used the strings with the names of the columns.