Reproducibility of research can be an overwhelming challenge for research labs of any discipline dealing with the complexity of modern software. Uber Elevate’s research arm ATCP has been working on a reproducible Research Platform based on Bazel allowing us to confidently explore new research ideas – while mitigating technical debt, incrementally and sustainably build upon previous work, create scalable experimental designs and build reproducible research artifacts.
Our team shares this methodology at the intersection of modern software and research engineering, our recent contributions to the Bazel ecosystem as well as lessons learnt as we battled through difficulties along the way. We also share our vision as we plan out the next steps: share our work to everybody interested and keep extending our use of Bazel as a generic build tool beyond software artifacts.
Resources:
Research on Reproducibility crisis: https://goo.gle/3ll1u6p
Introducing Maneage: https://goo.gle/32tNQGU
Artificial intelligence faces reproducibility crisis: https://goo.gle/35hGjN1
Speakers: Sebastien Mamessier & Romain Pennec
#bazelcon #bazel event: Bazelcon 2020; re_ty: Publish; product: Open Source – General; fullname: Sebastien Mamessier, Romain Pennec;"
