Molecular communication uses molecules or other nanoscale particles to transmit data in scenarios where conventional communication techniques are not feasible. In previous work a testbed using superparamagnetic iron oxide nanoparticles (SPIONs) as information carriers in a fluid transmission channel with constant background flow was proposed. The SPIONs are detected at a receiver as change of a coils inductance. We now improve the testbed by using a piezoelectric micropump as transmitter, making amplitude modulation (AM) with di?erent injection volumes possible. Machine learning is employed at the receiver to di?erentiate between six di?erent amplitude levels and grey code is used to reduce bit errors. With AM and the designed coding scheme, the achievable e?ective data rate was doubled to 4.45 bit/s.