While there are many job databases geared towards matching acronyms in my resume with those in an employer's ads, there don't seem to be any that keep track of my single most important factor in a choice of job - who already works there.
In my fictitious database, users sign up, get an ID, and submit
(and keep updated) four things:
(a) A resume.
(b) A selection from a set list of keywords that describes the person's expertise.
(c) A list of people they never want to work with again.
(d) A list of people they would enjoy working with.
Based on this information, scientists can begin collaborations if they like each other's work, without having to go through the usual courtship of conferences. If they like each other, they're told; if they're not told, it may be because of one of them not being in the system, rather than of outright dislike, so there's no direct rejection.
Based on this information, employers can find stable groups of people, not just single people, that work well together, or are interested in each other. Answering a request like "I'd like a four-person research department with a tendency towards AI and Human Interfaces" becomes solveable by a simple matchmaking algorithm.
(Problem: the data in such a database would be so precious that I'm not sure anyone would trust a company with it; realistic opinions about how desirable a co-worker an employee is would be used to evaluate employees, not just find matches.)