Export single pandas dataframe to multiple SQL tables (automatic normalization)

I have a DataFrame like this, but with millions of rows and about 15 columns:

       id    name  col1   col2  total
0 8252552 CHARLIE DESC1 VALUE1   5.99
1 8252552 CHARLIE DESC1 VALUE2  20.00
2 5699881    JOHN DESC1 VALUE1  39.00
2 5699881    JOHN DESC2 VALUE3  -3.99

The DataFrame needs to be exported to a SQL database, in several tables. I'm currently using SQLite3, to test the functionality. The tables would be:

• main (id INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT UNIQUE, people_id INTEGER, col1_id INTEGER, col2_id INTEGER, total REAL)
• people (id INTEGER NOT NULL PRIMARY KEY UNIQUE, name TEXT UNIQUE)
• col1 (id INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT UNIQUE, name TEXT UNIQUE)
• col2 (id INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT UNIQUE, name TEXT UNIQUE)

The main table should look similar to this:

  people_id col1_id col2_id  total
0   8252552       1       1   5.99
1   8252552       1       2  20.00
2   5699881       1       1  39.00
3   5699881       2       3  -3.99

Other tables, like "people", like this:

     id    name
8252552 CHARLIE
5699881    JOHN

Thing is, I can't find how to achieve that using the schema attribute of the to_sql method in pandas. Using Python, I'd do something like this:

conn = sqlite3.connect("main.db")
cur = conn.cursor()
for row in dataframe:
    id = row["ID"]
    name = row["Name"]
    col1 = row["col1"]
    col2 = row["col2"]
    total = row["total"]
    cur.execute("INSERT OR IGNORE INTO people (id, name) VALUES (?, ?)", (id, name))
    people_id = cur.fetchone()[0]
    cur.execute("INSERT OR IGNORE INTO col1 (col1) VALUES (?)", (col1, ))
    col1_id = cur.fetchone()[0]
    cur.execute("INSERT OR IGNORE INTO col1 (col2) VALUES (?)", (col2, ))
    col2_id = cur.fetchone()[0]
    cur.execute("INSERT OR REPLACE INTO main (people_id, col1_id, col2_id, total) VALUES (?, ?, ?, ?)", (people_id, col1_id, col2_id, total ))
conn.commit()

That would automatically add the corresponding values to the tables (people, col1 and col2), create a row with the desire values and foreign keys, and add that row to the primary table. However, there are a lot of columns and rows, and this might get very slow. Plus, I don't feel very confident that this is a "best practice" when dealing with databases (I'm fairly new to database development)

My question is: Is there a way to export a pandas DataFrame to multiple SQL Tables, setting the normalization rules, as in the above example? Is there any way to get the same result, with improved performance?