How RnD Appears in Large Tech Companies

Search growth required processing increasing volumes of information and generated infrastructure challenges. Google's hybrid approach is presented as a portfolio spanning fundamental work, new products, contributions to existing products, and infrastructure. Projects can have different time horizons and risks rather than conforming to one universal research-center model.

Engineering examples include GFS, MapReduce, Bigtable, Spanner, Borg, and other systems. Their papers influenced later open implementations, but publication often followed development. The author also distinguishes publishing an idea from releasing an accessible tool: early Google work frequently inspired external equivalents, while Kubernetes and TensorFlow illustrate direct participation in an open ecosystem.

Amazon's history is connected with e-commerce, its marketplace for sellers, and cloud infrastructure. Its research approach works backward from customer needs. Compute, storage, and database services are particularly relevant to engineering R&D: they address large technical challenges while also becoming products for external users.

The author contrasts ways of distributing results. In the early period discussed, Amazon more often released an infrastructure service than a detailed research paper. Later, a substantial publication stream emerged, including work on Aurora and DynamoDB. The comparison shows that research outcomes can take the form of products and interfaces on which other companies rely, not only papers. It is an interpretation of company history rather than a ranking of research organizations.

At Tinkoff, customer growth exposed constraints in commercial software, external services, and some open-source tools. The concept described combines internal teams with universities to address difficult, long-term problems. Its objectives are to support business scaling, create durable advantages, and attract strong engineers. Establishing a separate department does not by itself demonstrate that these objectives have been achieved.

The areas named include engineering productivity, distributed-system reliability, optimization, cloud systems, and security. Examples of outcomes include Voice Kit, Sage, ETNA, and Data Detective. These make the connection between research and company needs concrete. Applying the experience elsewhere starts with identifying where existing solutions constrain development and how results will be used, before selecting the organizational form of research and collaboration.