But to encourage innovative thinking you often (paradoxically) need to slow down.
You can't adopt a new technology without having deep insight and practical experience and we apply explorative, constructive learning approaches to gain them.
We define small projects, knowing that the results cannot be reused directly. They pay pay back in knowledge.
Topics were: symbolic parallelism on synthetic cases, leading to a remote benchmark test at the HPC centre in Cambridge (how far can you go with linear speed-up); combination of analytic and data driven methods with the vision of surrogate models for highest performance requirements; principle component application in simulation; Montecarlo and least square methods; ... all with comprehensive testing.
And we believe in learning-by-teaching. It really makes a difference if you give a talk on a conference or teach a 15 days course with live examples and immediate questions from experienced practitioners.
Pace at value needs creative slow downs from time to time.