Google's Hybrid Approach to Research
10 Dec 2016
Notes on the article by Spector, Norvig, and Petrov on the way google organizes the combination of engineering and research.
empirical component of CS research is in upswing. But CS in general is expanding. Methods of dissemination is evolving, from only papers, to e.g. open source releases.
goal of research is to bring significant, practical benefits to the users, and to do so rapidly.
research and development are closely integrated, in fact line between the two is very blurred.
There are research teams where the emphasis is more on research, but the distinction is one of emphasis. People and teams can move between the two. This integration minimizes or eliminates the issues with transfer from research to implementation.
Long term research is factorized into smaller steps (where possible; there are things where long term is essential, but these are the exception).
Iterative approach; avoiding elaborate prototypes, immediately going into production where possible. Teams move with product from research to production.
- aspects at google that support this:
- there is a rich set of computing abstractions and resources available for universal use at google
- the services based model make this all at everyones disposal
- the teams across the organization consist of talented individuals
being part of research community reduces the probability of being blindsided by developments.
- risk of the integration is that there is not enough openness for academic developments; mitigate this by very open exchange with academic community.