10 years ago I worked with a few coworkers on a protein search algorithm through the PDB (finding structure motifs with flexible and more stable components). It was a multi-month effort and our tool eventually vanished into the failed venture backed bio company. This type of problem is trivial compared to energy minimization (simulated structure) efforts.
Mapping problems to an interface where thousands or millions of humans and computers can tackle them generates a sense of awe each time I read about it.
I was in a similar situation I worked at a startup (which also went under) writing algos to predict protein structures given a minimal set of information (a very hard problem to solve and computationally expensive). Rosetta first showed up on the scene at that time and was a real shot in the arm for the field. One could argue that the game mechanics of Foldit are the most interesting thing we've seen in the structure prediction field in a long time.
Mapping problems to an interface where thousands or millions of humans and computers can tackle them generates a sense of awe each time I read about it.