Note that the Q factor of those will be quite poor. There's a lot of choke points (inner turn edges, vias) and the faces are broad, no stranding or whatever to avoid eddy currents (not that litz works very well at 27MHz, anyway). That's on top of the highly unfavorable geometry -- the narrow aspect ratio leaves very little field volume, for all that resistance (a square cross section is preferable).
It could be improved simply by using a thicker board (1/8" say), which could even lead to some other optimizations like putting those relatively bulky ceramic caps down into the board -- probably not as directly embedded components (some things can actually be fabbed in like this, but I doubt components of that size can?), but it would be fine to rout pockets for them, and do something goofy with castellated holes to allow wave soldering after gluing them in, perhaps. Maybe with some strain relief so they don't get torn apart / snapped by board flex, at even more expense to stray inductance of course (which is already lower due to not having a ground plane under the body -- connection can only be made edgewise off the side of the component). Such a build would have the unique distinction of being essentially flat; well, the transistors and support components will still stick out, but those can be almost arbitrarily small if we allow die bonding say.
On the upside, Q generally goes as sqrt(F), so for the frequency being 100 times the usual switching range, it might go from a uselessly low Q, to a passable if inefficient (say 30s?) level.
On a related note, Murata makes a line of DC-DC converters with embedded ferrite core transformers, "wound" in much the same way. They're made as a sandwich I think, gluing together three panels, or perhaps two with a milled pocket in one. So there's no problem with prepreg squidging out around the embedded component, it's dead air space inside rigid materials. Then the stack is drilled and plated, and there's your vias.
Tim