How to create tensor from a csv file in onnx runtime?

Question

I've trained a CNN model on python, and saved it as an onnx model. Now I want to use it in a c# winforms application, to perform inference on .csv files. I read that one should use a onnx runtime for this, since it's great for performance, and I'll want to perform real-time inferences eventually. Problem is, I'm confused as to how to convert my .csv files into tensors in c# & onnx runtime?

I'm more familiar with python, and i've managed to convert it there no problem:

def retrieveCSV(filepath, variable, minrange, maxrange):
    df = pd.DataFrame()
    for file in filenames:
        orig_df = pd.read_csv(file, index_col = False, sep=',')
        var_df = orig_df[[variable]].reset_index() 
        var_df= var_df.iloc[minrange : maxrange].T
        df = pd.concat([df, var_df.iloc[1:]], ignore_index=True, sort=False, axis = 0)\
    df= pd.DataFrame(MinMaxScaler().fit_transform(df))
    return df

I have converted df to a tensor after that; just giving some context as to what I'm trying to now replicate in c#. Am wondering if it'll be easier if I just used some sort of wrapper like Python.NET, to port my python functions over to c#?