From here:
For the first time, Tufts University biologists have reported that bioelectrical signals are necessary for normal head and facial formation in an organism and have captured that process in a time-lapse video that reveals never-before-seen patterns of visible bioelectrical signals outlining where eyes, nose, mouth, and other features will appear in an embryonic tadpole.
The Tufts biologists found that, before the face of a tadpole develops, bioelectrical signals (ion flux) cause groups of cells to form patterns marked by different membrane voltage and pH levels. When stained with a reporter dye, hyperpolarized (negatively charged) areas shine brightly, while other areas appear darker, creating an “electric face.”
“When a frog embryo is just developing, before it gets a face, a pattern for that face lights up on the surface of the embryo,” said senior author Dany S. Adams, Ph.D. Adams is a research associate professor in the Department of Biology in the Tufts School of Arts and Sciences and a member of the Tufts Center for Regenerative and Developmental Biology. “We believe this is the first time such patterning has been reported for an entire structure, not just for a single organ. I would never have predicted anything like it. It’s a jaw dropper.”

So do we know what causes the bioelectrical signals?
Bilbo,
From the news article:
Tufts Post Doctoral Associate Laura N. Vandenberg, Ph.D., was first author of the paper entitled “V-ATPase-dependent ectodermal voltage and pH regionalization are required for craniofacial morphogenesis.” Ryan D. Morrie, a biology major in the School of Arts and Sciences, was second author.
So, to answer your question – molecular machines.