As the waveform has sub-ms features, you need to sample with a much smaller period, say 10kHz, to see detail around the peak.
If you are stuck with only sampling at 100Hz, you would need to lowpass filter to less than 50Hz, which would destroy all the detail of the pulses, in particular the peak heights. However, you could make some measurements with your oscilloscope that could allow you to relate the proper measurement you have already made to a measurement through a lowpass filter, making some assumptions that the pulse shape and frequency stays the same.
Does the DAQ have a lowpass filter on its input, or does it sample? If it samples, then you can get stochastic information from simply taking a lot of samples, as long as the sampling is uncorrelated with the signal pulses. You will lose the shape of the waveform, but you will learn what voltage value the waveform exceeds for 90% of its time, or 99% of its time, which might be a good proxy for the peak.