You have available 100Vpp or 50V square amplitude. Theoretically optimal filtering would extract surprisingly 63,7V peak sinusoidal voltage from it. It means 45V RMS.
The RC filtering is far from ideal. You get DC through it and you must attenuate the fundamental sinusoidal component to keep the higher harmonics invisible at least in the oscilloscope.
You should replace your RC filter by a narrow band bandpass filter.
You use the mosfet as cathode follower. Your max output voltage to the filter is about the same as at the gate of the mosfet. Another answer gives a fix: use the fet as common cathode switch. Beware: You will dissipate hefty 5 watts in R10 and have 50VDC at the ouput.
Consider to generate the sinusoidal in low voltage and have an amp or transformer or both.
Just in case you want to generate the final sinusoidal voltage without a transformer and you do not want an amp due the low current consumption, I give one possiblity:
The resonant circuit L1C1 is the simplest possible bandpass filter that has reasonably low bandwidth. It's excited by current limited mosfet that pulls 10 kHz square pulses. Very soon there builds up a remarkable AC voltage to the drain of the mosfet. The voltage swings approximately 0V to 2 x Vsupply. If you have 50 volts Vsupply, you can expect 100Vpp AC voltage.
There are losses, but the theoretical overshoot partially compensates that.
A proper L1 is easily available as HiFi speaker tweeter filter coil. You must connect C1 from several capacitors to get the right tuning despite the coarse tolerances of the coil and capacitors. Elko is not an usable capacitor type, have for example polycarbonate or polyester caps. Note: The voltage rating. You must also select a type that stands at 10 kHz that 100Vpp or whatever your voltage is. This requirement IS NOT TRIVIAL!!!! A 100V DC-only type will get internally hot and blow.
There's a simple highpass filter 100nF / 100kOhm that removes the DC. It has a time constant, so you do better if you add a pair of opposite zener diodes that prevent the overvoltage in the startup.
You can add different resistors as loads to see how the output voltage drops when loaded. A few microamperes mean nothing. Probably you can calculate how much energy is needed to accelerate a known viscous fluid particle to a wanted velocity and move it to a wanted distance. This is your real load.
You can add a buffer voltage follower amp for greater output currents and for isolating the filter from capacitive loads.
Consider to use high voltage solid state AC relays for distributing the AC voltage. An example: Panasonic AQW216.